ggml : block interleaving support for Q4_K quanti for AArch64

* new quanti: block_q4_kx4 with offline repack impl

* new quantize path: add NEON impl for ggml_quantize_mat_q8_K_4x8

* new gemv kernel: new ggml_gemv_q4_K_4x8_q8_K NEON kernel for GGML_OP_MUL_MAT_ID/GGML_OP_MUL_MAT

* new gemm kernel: new ggml_gemm_q4_K_4x8_q8_K NEON kernel for GGML_OP_MUL_MAT_ID/GGML_OP_MUL_MAT

* performance boost for both S_PP and S_TG

---------

Co-authored-by: yuanjia111 <[email protected]>

Author: hongyang-7

ggml/src/ggml-cpu/arch-fallback.h

@@ -50,7 +50,7 @@
 #define ggml_gemm_iq4_nl_8x8_q8_0_generic ggml_gemm_iq4_nl_8x8_q8_0
 #elif defined(__aarch64__) || defined(__arm__) || defined(_M_ARM) || defined(_M_ARM64)
 // repack.cpp
-#define ggml_quantize_mat_q8_K_4x8_generic ggml_quantize_mat_q8_K_4x8
+//#define ggml_quantize_mat_q8_K_4x8_generic ggml_quantize_mat_q8_K_4x8
 #define ggml_gemv_q4_K_8x8_q8_K_generic ggml_gemv_q4_K_8x8_q8_K
 #define ggml_gemv_iq4_nl_8x8_q8_0_generic ggml_gemv_iq4_nl_8x8_q8_0
 #define ggml_gemv_q2_K_8x8_q8_K_generic ggml_gemv_q2_K_8x8_q8_K