[GPU] Support different input and output data type in convolution ref (openvinotoolkit#26501)

### Details:
- For fp model, some convolutions may not be compressed to fp16
depending on the transformation policy and those convolutions may have
the fused node which is of fp16. Then convolution node input data type
will be fp32 while output data type fp16. Convolution needs to support
this case.

### Tickets:
 - 147689

Author: davidsnam-intel

src/plugins/intel_gpu/src/kernel_selector/kernels/convolution/convolution_kernel_ref.cpp

@@ -123,7 +123,8 @@ bool ConvolutionKernel_Ref::Validate(const Params& params) const {
 
     // int8/uint8 inputs (quantization case) require additional checks
     // require some additional checks.
-    if (input_type == output_type && input_type != Datatype::UINT8 && input_type != Datatype::INT8)
+    if (input_type != Datatype::UINT8 && input_type != Datatype::INT8 &&
+        output_type != Datatype::UINT8 && output_type != Datatype::INT8)
         return true;
 
     // (u)int8 input + fp weights

src/plugins/intel_gpu/tests/unit/test_cases/convolution_gpu_test.cpp

@@ -12,6 +12,7 @@
 #include <intel_gpu/primitives/crop.hpp>
 #include <intel_gpu/primitives/reorder.hpp>
 #include <intel_gpu/primitives/reshape.hpp>
+#include <intel_gpu/primitives/permute.hpp>
 
 #include <algorithm>
 #include <array>
@@ -1639,6 +1640,47 @@ TEST(convolution_f32_fw_gpu, basic_convolution) {
     }
 }
 
+TEST(convolution_f32_fw_gpu, input_f32_output_f16_dynamic_ref_kernel) {
+    auto& engine = get_test_engine();
+
+    auto in_layout = layout{ov::PartialShape::dynamic(4), data_types::f32, format::bfyx};
+    auto input = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 5, 4 } });
+    auto weights = engine.allocate_memory({ data_types::i8, format::bfyx, { 1, 1, 3, 2 } });
+    auto biases = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+    auto eltwise_data = engine.allocate_memory({ data_types::f32, format::bfyx, { 1, 1, 1, 1 } });
+
+    set_values(input, {
+        1.0f, 2.0f, 3.0f, 4.0f,
+        5.0f, 2.0f, 2.0f, 3.0f,
+        4.0f, 6.0f, 3.0f, 3.0f,
+        3.0f, 5.0f, 1.0f, 1.0f,
+        1.0f, 1.0f, 1.0f, 1.0f }
+    );
+
+    topology topology(
+        input_layout("input", in_layout),
+        data("weights", weights),
+        data("biases", biases),
+        data("eltwise_data", eltwise_data),
+        convolution( "conv", input_info("input"), "weights", "biases", 1, {2, 1}, {1, 1}, {0, 0}, {0, 0}, false),
+        eltwise("eltwise", { input_info("conv"), input_info("eltwise_data") }, eltwise_mode::prod, data_types::f16),
+        permute("permute", input_info("eltwise"), {0, 1, 2, 3}));
+
+    ExecutionConfig config = get_test_default_config(engine);
+    ov::intel_gpu::ImplementationDesc conv_impl_ref = { format::bfyx, "convolution_gpu_ref", impl_types::ocl };
+    config.set_property(ov::intel_gpu::force_implementations(ov::intel_gpu::ImplForcingMap{ { "conv", conv_impl_ref } }));
+    config.set_property(ov::intel_gpu::allow_new_shape_infer(true));
+
+    network network(engine, topology, config);
+    network.set_input_data("input", input);
+
+    auto outputs = network.execute();
+
+    ASSERT_EQ(outputs.size(), size_t(1));
+    ASSERT_FALSE(has_node(*network.get_program(), "eltwise"));
+    ASSERT_EQ(outputs.at("permute").get_layout().data_type, data_types::f16);
+}
+
 TEST(convolution_f32_fw_gpu, convolution_big_size_weights) {
     auto& engine = get_test_engine();