Revert "[SYCL] fallback mmvq (ggml/9088)" (llama/9579)

ggml : refactoring (llama/#0)

ggml : fix builds (llama/0)

sycl : update support conditions (llama/9394)

add check malloc result on device (llama/9346)

ggml/examples: add backend support for numerical optimization (ggml/949)

Add oneDNN primitive support (llama/9091)

fallback mmvq (llama/9088)

Fix SYCL `im2col` and `convert` Overflow with Large Dims (llama/9052)

Add `TIMESTEP_EMBEDDING` OP (llama/8707)

add conv support (llama/8688)

ggml : reduce hash table reset cost (llama/8698)

add concat through dim 1/2 (llama/8483)

fix the mul_mat_id ut issues (llama/8427)

sycl : Reenabled mmvq path for the SYCL Nvidia Backend (llama/8372)

Enabled more data types for oneMKL gemm_batch (llama/8236)

Fix WARP_SIZE=16 bug of Intel GPU (llama/8266)

rm get_work_group_size() by local cache for performance (llama/8286)

Fix the sub group size of Intel (llama/8106)

Update SYCL-Rope op and Refactor (llama/8157)

whisper : reorganize source code + improve CMake (#2256)