[CPU] Apply 'cppcoreguidelines-pro-type-cstyle-cast' clang-tidy remarks (openvinotoolkit#30037)

### Details:
- Fix "cppcoreguidelines-pro-type-cstyle-cast" remarks reported by
clang-tidy
- Enable "cppcoreguidelines-pro-type-cstyle-cast" clang-tidy checks on
CI by default

### Tickets:
 - N/A

Author: aobolensk

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

@@ -52,6 +52,7 @@ Checks: >
   modernize-*,
   misc-*,
   cppcoreguidelines-prefer-member-initializer,
+  cppcoreguidelines-pro-type-cstyle-cast,
   readability-container-size-empty,
   readability-else-after-return,
   -bugprone-easily-swappable-parameters,

src/plugins/intel_cpu/src/dnnl_postops_composer.cpp

@@ -940,8 +940,8 @@ void DnnlPostOpsComposer::appendAttrPostOpsLegacy(const FakeQuantizePostOp& post
         }
 
         return ops.append_binarization(dnnl::algorithm::binarization_depthwise,
-                                       (const float*)&binarizationThresholds[0],
-                                       (const float*)&binarizationOutputMask[0]);
+                                       reinterpret_cast<const float*>(binarizationThresholds.data()),
+                                       reinterpret_cast<const float*>(binarizationOutputMask.data()));
     }
 
     dnnl::algorithm alg = postOp.type() == FakeQuantizePostOp::Type::quantization_only

src/plugins/intel_cpu/src/emitters/snippets/x64/kernel_executors/brgemm_copy_b.cpp

@@ -164,7 +164,7 @@ BrgemmCopyBKernel::BrgemmCopyBKernel(const BrgemmCopyBKernelConfig& conf)
 status_t BrgemmCopyBKernel::create_kernel() {
     const auto code = jit_generator::create_kernel();
     OV_CPU_JIT_EMITTER_ASSERT(code == status::success, "Failed to create kernel");
-    ker_ = (decltype(ker_))jit_ker();
+    ker_ = reinterpret_cast<decltype(ker_)>(const_cast<uint8_t*>(jit_ker()));
     return code;
 }
 

src/plugins/intel_cpu/src/nodes/color_convert.cpp

@@ -127,7 +127,7 @@ void jit_uni_converter::init() {
     if (create_kernel() != status::success) {
         OPENVINO_THROW("Can't generate jit color converter kernel");
     }
-    _fn = (function_t)jit_ker();
+    _fn = reinterpret_cast<function_t>(const_cast<uint8_t*>(jit_ker()));
 }
 
 template <size_t N>

src/plugins/intel_cpu/src/nodes/input.cpp

@@ -37,7 +37,9 @@ struct jit_has_special_value_base : public jit_generator {
     }
 
     fn_t get() {
-        return jit_ker() || create_kernel() == dnnl::impl::status::success ? (fn_t)jit_ker() : nullptr;
+        return jit_ker() || create_kernel() == dnnl::impl::status::success
+                   ? reinterpret_cast<fn_t>(const_cast<uint8_t*>(jit_ker()))
+                   : nullptr;
     }
 
 protected:
@@ -199,9 +201,9 @@ struct jit_has_subnormals : public jit_has_special_value_base {
 
         // Initialize necessary consts
         uni_vpxor(zero, zero, zero);
-        mov(reg_mask_addr, (size_t)exponent_mask_data);
+        mov(reg_mask_addr, reinterpret_cast<size_t>(exponent_mask_data));
         uni_vmovdqu(exponent_mask, ptr[reg_mask_addr]);
-        mov(reg_mask_addr, (size_t)mantissa_mask_data);
+        mov(reg_mask_addr, reinterpret_cast<size_t>(mantissa_mask_data));
         uni_vmovdqu(mantissa_mask, ptr[reg_mask_addr]);
 
         // Main loop
@@ -273,9 +275,9 @@ struct jit_has_bf16_overflows : public jit_has_special_value_base {
 
         // Initialize necessary consts
         uni_vpxor(zero, zero, zero);
-        mov(reg_mask_addr, (size_t)bf16_max_mask_data);
+        mov(reg_mask_addr, reinterpret_cast<size_t>(bf16_max_mask_data));
         uni_vmovdqu(bf16_max_mask, ptr[reg_mask_addr]);
-        mov(reg_mask_addr, (size_t)bf16_min_mask_data);
+        mov(reg_mask_addr, reinterpret_cast<size_t>(bf16_min_mask_data));
         uni_vmovdqu(bf16_min_mask, ptr[reg_mask_addr]);
 
         // Main loop
@@ -417,10 +419,11 @@ void Input::cloneBlobIfRequired() {
                 std::atomic<bool> has_bf16_overflows_local(false);
                 if (needFlushDenormalsToZero || do_bf16_saturation_check) {
                     parallel_for(iterations_num, [&](int n) {
-                        auto ptr = u32data + n * batch_size;
-                        jit_has_special_value_base::args_t args = {reinterpret_cast<float const*>(ptr),
-                                                                   std::min(batch_size, (size_t)(u32data + size - ptr)),
-                                                                   false};
+                        auto ptr = f32data + n * batch_size;
+                        jit_has_special_value_base::args_t args = {
+                            reinterpret_cast<const float*>(ptr),
+                            std::min(batch_size, static_cast<size_t>(f32data + size - ptr)),
+                            false};
 
                         if (needFlushDenormalsToZero && !has_subnormals_local) {
                             fn(&args);