[CPU] Apply 'modernize-*' clang-tidy remarks (openvinotoolkit#28804)

### Details:
 - Fix "modernize-*" remarks reported by clang-tidy
 - Enable "modernize-*" clang-tidy checks on CI by default
- Disable "modernize-avoid-c-arrays" due to big scope of changes (to be
enabled separately)
- Disable C++20 specific checks: "modernize-use-constraints",
"modernize-use-std-numbers"

### Tickets:
 - N/A

Author: aobolensk

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

@@ -27,16 +27,19 @@
 # -google-default-arguments,
 # -google-explicit-constructor,
 # -google-readability-casting,
+# -modernize-avoid-c-arrays,
 # -readability-implicit-bool-conversion,
 # -readability-magic-numbers, cppcoreguidelines-avoid-magic-numbers
 # -readability-function-cognitive-complexity. Reasonable way to enforce splitting complex code into simple functions
+# Remove warning disablement after CI pipeline migrates to C++17 from C++20 for:
+# -modernize-use-constraints,
+# -modernize-use-std-numbers
 
 Checks: >
   -*,
   performance-*,
   google-*,
-  modernize-loop-convert,
-  modernize-pass-by-value,
+  modernize-*,
   cppcoreguidelines-prefer-member-initializer,
   readability-else-after-return,
   -bugprone-easily-swappable-parameters,
@@ -49,6 +52,9 @@ Checks: >
   -google-readability-casting,
   -google-readability-todo,
   -readability-braces-around-statements,
+  -modernize-avoid-c-arrays,
+  -modernize-use-constraints,
+  -modernize-use-std-numbers,
   -modernize-use-trailing-return-type,
   -readability-identifier-length,
   -readability-implicit-bool-conversion,
@@ -74,6 +80,8 @@ CheckOptions:
     value: reasonable
   - key: modernize-pass-by-value.IncludeStyle
     value: google
+  - key: modernize-use-auto.MinTypeNameLength
+    value: "3"
 ### To be considered to enable:
 #  # Unifies the usage of the statements
 #  - key: readability-braces-around-statements.ShortStatementLines

src/plugins/intel_cpu/src/cpu_memory.cpp

@@ -32,8 +32,8 @@ BlockedMemoryDescPtr IMemory::getDescWithType<BlockedMemoryDesc, 0, 0>() const {
 
 namespace {
 inline void setSubnormalsToZeroAndbf16Saturation(float* data, size_t size, bool ftz, bool bf16saturation) {
-    uint32_t* u32data = reinterpret_cast<uint32_t*>(data);
-    float* floatdata = reinterpret_cast<float*>(data);
+    auto* u32data = reinterpret_cast<uint32_t*>(data);
+    auto* floatdata = reinterpret_cast<float*>(data);
     if (ftz && bf16saturation) {
         for (size_t i = 0; i < size; ++i) {
             if ((u32data[i] & (0xFF << 23)) == 0) {
@@ -583,7 +583,7 @@ bool mbind_move(void* data, size_t size, int targetNode) {
     int realNode = ov::get_org_numa_id(targetNode);
     auto pagesize = getpagesize();
     auto page_count = (size + pagesize - 1) / pagesize;
-    char* pages = reinterpret_cast<char*>(  // NOLINT(performance-no-int-to-ptr)
+    auto* pages = reinterpret_cast<char*>(  // NOLINT(performance-no-int-to-ptr)
         ((reinterpret_cast<uintptr_t>(data)) & ~(static_cast<uintptr_t>(pagesize - 1))));
     uint64_t mask = 0;
     unsigned flags = 0;
@@ -663,7 +663,7 @@ MemoryPtr split_horizontal(const dnnl::engine& eng,
     VectorDims stride_dims = dims;
     stride_dims[dim] = splited_dim_vec[0];
     size_t stride =
-        std::accumulate(stride_dims.begin(), stride_dims.end(), static_cast<size_t>(1), std::multiplies<size_t>()) *
+        std::accumulate(stride_dims.begin(), stride_dims.end(), static_cast<size_t>(1), std::multiplies<>()) *
         prec.size();
 
     // create new shape for target memory

src/plugins/intel_cpu/src/cpu_streams_calculation.cpp

@@ -24,8 +24,8 @@
 using namespace ov;
 using namespace ov::threading;
 
-#define INIT_VAL     -100
-#define TP_CPU_LIMIT 32
+constexpr int INIT_VAL = -100;
+constexpr int TP_CPU_LIMIT = 32;
 
 namespace ov::intel_cpu {
 
@@ -165,7 +165,7 @@ std::vector<std::vector<int>> get_streams_info_table(
 
     auto check_threads_per_stream = [&]() {
         int count = 0;
-        while (1) {
+        while (true) {
             for (int n_type = MAIN_CORE_PROC; n_type <= HYPER_THREADING_PROC; n_type++) {
                 count += static_cast<int>(proc_type_table[0][n_type] / n_threads_per_stream);
             }

src/plugins/intel_cpu/src/dnnl_extension_utils.cpp

@@ -148,7 +148,7 @@ VectorDims DnnlExtensionUtils::convertToVectorDims(const memory::dims& dims) {
 }
 
 VectorDims DnnlExtensionUtils::convertToVectorDims(const dnnl::impl::dims_t dims, const int ndims) {
-    return VectorDims(dims, dims + ndims);
+    return {dims, dims + ndims};
 }
 
 memory::dims DnnlExtensionUtils::convertToDnnlDims(const VectorDims& dims) {
@@ -227,7 +227,7 @@ std::string DnnlExtensionUtils::query_impl_info_str(const const_dnnl_primitive_d
     if (status != dnnl_success) {
         OPENVINO_THROW("query_impl_info_str failed.");
     }
-    return std::string(res);
+    return res;
 }
 
 bool DnnlExtensionUtils::find_implementation(dnnl::primitive_desc& desc, impl_desc_type impl_type) {

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

@@ -242,8 +242,8 @@ template <cpu_isa_t isa>
 void jit_convert_truncation_emitter::emit_isa(const std::vector<size_t>& in_idxs,
                                               const std::vector<size_t>& out_idxs) const {
     using TReg = typename dnnl::impl::cpu::aarch64::cpu_isa_traits<isa>::TReg;
-    TReg src = TReg(in_idxs[0]);
-    TReg dst = TReg(out_idxs[0]);
+    auto src = TReg(in_idxs[0]);
+    auto dst = TReg(out_idxs[0]);
     jit_convert_process<TReg>(src, dst, input_type, output_type, false);
 }
 
@@ -267,8 +267,8 @@ template <cpu_isa_t isa>
 void jit_convert_saturation_emitter::emit_isa(const std::vector<size_t>& in_idxs,
                                               const std::vector<size_t>& out_idxs) const {
     using TReg = typename dnnl::impl::cpu::aarch64::cpu_isa_traits<isa>::TReg;
-    TReg src = TReg(in_idxs[0]);
-    TReg dst = TReg(out_idxs[0]);
+    auto src = TReg(in_idxs[0]);
+    auto dst = TReg(out_idxs[0]);
     jit_convert_process<TReg>(src, dst, input_type, output_type, true);
 }