[CPU] Fix modernize-loop-convert clang-tidy remarks (#28966)

### Details:
 - Fix "modernize-loop-convert" remarks reported by clang-tidy
 - Enable "modernize-loop-convert" clang-tidy checks on CI by default

### Tickets:
 - N/A

Author: aobolensk

src/plugins/intel_cpu/src/.clang-tidy

@@ -35,6 +35,7 @@ Checks: >
   -*,
   performance-*,
   google-*,
+  modernize-loop-convert,
   modernize-pass-by-value,
   cppcoreguidelines-prefer-member-initializer,
   -bugprone-easily-swappable-parameters,

src/plugins/intel_cpu/src/cpu_streams_calculation.cpp

@@ -212,9 +212,9 @@ std::vector<std::vector<int>> get_streams_info_table(
         ((input_streams_changed == true) && (input_streams == 1))) {
         n_streams = 1;
         stream_info[NUMBER_OF_STREAMS] = n_streams;
-        for (size_t n = 0; n < proc_socket_table.size(); n++) {
-            if (proc_socket_table[n][ALL_PROC] > 0) {
-                current_socket_id = proc_socket_table[n][PROC_SOCKET_ID];
+        for (auto& n : proc_socket_table) {
+            if (n[ALL_PROC] > 0) {
+                current_socket_id = n[PROC_SOCKET_ID];
                 break;
             }
         }

src/plugins/intel_cpu/src/dnnl_postops_composer.cpp

@@ -527,8 +527,8 @@ bool DnnlPostOpsComposer::appendScale(const std::vector<float>& scale, bool isLa
                 wei_scale_values[j] *= scale[j];
             }
         } else {
-            for (size_t j = 0; j < wei_scale_values.size(); j++) {
-                wei_scale_values[j] *= scale[0];
+            for (float& wei_scale_value : wei_scale_values) {
+                wei_scale_value *= scale[0];
             }
         }
 

src/plugins/intel_cpu/src/dnnl_postops_composer_legacy.cpp

@@ -169,8 +169,8 @@ bool DnnlPostOpsComposerLegacy::appendScale(const std::vector<float>& scale, boo
                 wei_scale_values[j] *= scale[j];
             }
         } else {
-            for (size_t j = 0; j < wei_scale_values.size(); j++) {
-                wei_scale_values[j] *= scale[0];
+            for (float& wei_scale_value : wei_scale_values) {
+                wei_scale_value *= scale[0];
             }
         }
 

src/plugins/intel_cpu/src/emitters/plugin/aarch64/jit_emitter.cpp

@@ -62,8 +62,8 @@ void jit_emitter::emit_data() const {
     assert(sizeof(table_entry_val_t) == 4);
 
     // Run through the map and insert values stored there
-    for (auto it = entry_map_.begin(); it != entry_map_.end(); it++) {
-        const auto& te = (*it).second;  // get map entry for a given key
+    for (const auto& it : entry_map_) {
+        const auto& te = it.second;  // get map entry for a given key
         const auto len = te.bcast ? get_vec_length() : sizeof(table_entry_val_t);
         for (size_t d = 0; d < len; d += sizeof(table_entry_val_t)) {
             h->dd(te.val);
@@ -103,8 +103,8 @@ void jit_emitter::prepare_table() {
     // expect the same order when injecting the table entries in
     // prepare_table.
     size_t off = 0;
-    for (auto it = entry_map_.begin(); it != entry_map_.end(); it++) {
-        auto& te = (*it).second;
+    for (auto& it : entry_map_) {
+        auto& te = it.second;
         te.off = off;
         off += te.bcast ? get_vec_length() : sizeof(table_entry_val_t);
     }

src/plugins/intel_cpu/src/emitters/plugin/aarch64/jit_emitter.hpp

@@ -141,9 +141,9 @@ class jit_emitter : public ov::snippets::Emitter {
     }
 
     void push_entries_of(const table_t& t) {
-        for (auto it = t.begin(); it != t.end(); it++) {
-            auto key = (*it).first;
-            auto te = (*it).second;  // copy values from table
+        for (const auto& it : t) {
+            auto key = it.first;
+            auto te = it.second;  // copy values from table
             push_arg_entry_of(key, te.val, te.bcast);
         }
     }

src/plugins/intel_cpu/src/emitters/plugin/x64/jit_emitter.cpp

@@ -93,11 +93,11 @@ void jit_emitter::emitter_preamble(const std::vector<size_t>& in_idxs,
         }
 
         // moving mask vector at the beginning of aux vectors list to simplify further processing
-        for (size_t i = 0; i < aux_vec_idxs.size(); i++) {
-            if (aux_vec_idxs[i] == 0) {
+        for (size_t& aux_vec_idx : aux_vec_idxs) {
+            if (aux_vec_idx == 0) {
                 size_t tmp = aux_vec_idxs[0];
-                aux_vec_idxs[0] = aux_vec_idxs[i];
-                aux_vec_idxs[i] = tmp;
+                aux_vec_idxs[0] = aux_vec_idx;
+                aux_vec_idx = tmp;
                 break;
             }
         }
@@ -176,8 +176,8 @@ void jit_emitter::emitter_preamble(const std::vector<size_t>& in_idxs,
         aux_gpr_idxs.erase(aux_gpr_idxs.end() - 1);
     }
 
-    for (size_t i = 0; i < preserved_gpr_idxs.size(); ++i) {
-        h->push(Reg64(preserved_gpr_idxs[i]));
+    for (uint64_t preserved_gpr_idx : preserved_gpr_idxs) {
+        h->push(Reg64(preserved_gpr_idx));
     }
 
     if (preserved_vec_idxs.size()) {
@@ -223,8 +223,8 @@ void jit_emitter::emit_data() const {
     assert(sizeof(table_entry_val_t) == 4);
 
     // Run through the map and insert values stored there
-    for (auto it = entry_map_.begin(); it != entry_map_.end(); it++) {
-        const auto& te = (*it).second;  // get map entry for a given key
+    for (const auto& it : entry_map_) {
+        const auto& te = it.second;  // get map entry for a given key
         const auto len = te.bcast ? get_vec_length() : sizeof(table_entry_val_t);
         for (size_t d = 0; d < len; d += sizeof(table_entry_val_t)) {
             h->dd(te.val);
@@ -240,8 +240,8 @@ void jit_emitter::prepare_table() {
     // expect the same order when injecting the table entries in
     // prepare_table.
     size_t off = 0;
-    for (auto it = entry_map_.begin(); it != entry_map_.end(); it++) {
-        auto& te = (*it).second;
+    for (auto& it : entry_map_) {
+        auto& te = it.second;
         te.off = off;
         off += te.bcast ? get_vec_length() : sizeof(table_entry_val_t);
     }

src/plugins/intel_cpu/src/emitters/plugin/x64/jit_emitter.hpp

@@ -152,9 +152,9 @@ class jit_emitter : public ov::snippets::Emitter {
     }
 
     void push_entries_of(const table_t& t) {
-        for (auto it = t.begin(); it != t.end(); it++) {
-            auto key = (*it).first;
-            auto te = (*it).second;  // copy values from table
+        for (const auto& it : t) {
+            auto key = it.first;
+            auto te = it.second;  // copy values from table
             push_arg_entry_of(key, te.val, te.bcast);
         }
     }

src/plugins/intel_cpu/src/graph.cpp

@@ -241,13 +241,13 @@ void Graph::Replicate(const std::shared_ptr<const ov::Model>& model,
         const auto& inputNode = input.second;
         const auto precToSet = inputNode->getOriginalOutputPrecisionAtPort(0);
         const auto childEdges = inputNode->getChildEdgesAtPort(0);
-        for (size_t i = 0; i < childEdges.size(); i++) {
-            const auto child = childEdges[i]->getChild();
-            const auto child_prec = child->getOriginalInputPrecisionAtPort(childEdges[i]->getOutputNum());
+        for (const auto& childEdge : childEdges) {
+            const auto child = childEdge->getChild();
+            const auto child_prec = child->getOriginalInputPrecisionAtPort(childEdge->getOutputNum());
             if (!one_of(child_prec, ov::element::bf16, ov::element::f16) &&
                 // remove this WA when #78939 is resolved
                 !hasSubgraphConsumers(child)) {
-                child->setOriginalInputPrecisionAtPort(childEdges[i]->getOutputNum(), precToSet);
+                child->setOriginalInputPrecisionAtPort(childEdge->getOutputNum(), precToSet);
             }
         }
     }
@@ -631,9 +631,9 @@ void Graph::ResolveEdgeConflicts() {
 
     /* When inserting convert / reorder, two new edges are added (pushed to the end) to the graphEdges.
        So use a plain for loop, to handle newly inserted edges as well */
-    for (size_t i = 0; i < graphEdges.size(); i++) {
+    for (size_t i = 0; i < graphEdges.size(); i++) {  // NOLINT(modernize-loop-convert)
         auto& edge = graphEdges[i];
-        auto reorderStatus = edge->needReorder();
+        auto reorderStatus = edge->needReorder();  // NOLINT(modernize-loop-convert)
         DEBUG_LOG(*edge, " reorderStatus = ", reorderStatus);
 
         switch (reorderStatus) {
@@ -1757,11 +1757,11 @@ void Graph::GetPerfData(std::vector<ov::ProfilingInfo>& perfMap) const {
             }
         };
 
-    for (size_t i = 0; i < graphNodes.size(); i++) {
-        if (graphNodes[i]->isConstant()) {
+    for (const auto& graphNode : graphNodes) {
+        if (graphNode->isConstant()) {
             continue;
         }
-        getPerfMapFor(perfMap, graphNodes[i]);
+        getPerfMapFor(perfMap, graphNode);
     }
 }
 
@@ -1793,7 +1793,8 @@ void Graph::DropNode(const NodePtr& node) {
     auto children = node->childEdges;
     auto parents = node->parentEdges;
 
-    for (size_t i = 0; i < parents.size(); i++) {
+    // The collections are being updated while iterating. So, range based for loops cannot be used.
+    for (size_t i = 0; i < parents.size(); i++) {  // NOLINT(modernize-loop-convert)
         auto p_edge = parents[i].lock();
         if (!p_edge) {
             continue;
@@ -1806,7 +1807,8 @@ void Graph::DropNode(const NodePtr& node) {
         const int inNum = p_edge->getInputNum();
         RemoveEdge(p_edge);
 
-        for (size_t j = 0; j < children.size(); j++) {
+        // The collections are being updated while iterating. So, range based for loops cannot be used.
+        for (size_t j = 0; j < children.size(); j++) {  // NOLINT(modernize-loop-convert)
             auto c_edge = children[j].lock();
             if (!c_edge) {
                 continue;
@@ -1851,7 +1853,8 @@ void Graph::DropDWConvNode(const NodePtr& node) {
         const int inNum = p_edge->getInputNum();
         RemoveEdge(p_edge);
 
-        for (size_t j = 0; j < children.size(); j++) {
+        // The collections are being updated while iterating. So, range based for loops cannot be used.
+        for (size_t j = 0; j < children.size(); j++) {  // NOLINT(modernize-loop-convert)
             auto c_edge = children[j].lock();
             if (!c_edge) {
                 continue;