Revert "x64: matmul: blocking: update for 4k LDA"

This reverts commit 27b519d due to performance regression in E2E testing

Author: ankalinin

src/cpu/x64/matmul/brgemm_matmul_utils.cpp

@@ -1672,9 +1672,7 @@ float matmul_amx_blocking_params_t::get_thread_balance_scores() {
 // returns score for current blocking parameters' values in range [0, 1]
 // for copied data reusage
 float matmul_amx_blocking_params_t::get_copied_data_reusage_scores() {
-    // Values based on measured performance
-    const bool is_lda_4k = (current_lda_ * a_dt_sz) % 4096 == 0;
-    const dim_t effective_m_chunk_sz = 64 * (is_lda_4k ? 1 : 4);
+    const dim_t effective_m_chunk_sz = 64 * 4;
     const dim_t desired_M_chunk_size = is_runtime_M
             ? effective_m_chunk_sz
             : nstl::min(M, effective_m_chunk_sz);