CUDA: app option to compile without FlashAttention (llama/12025)

MUSA: support ARM64 and enable dp4a .etc (llama/11843)

CUDA: use mma PTX instructions for FlashAttention (llama/11583)

ggml : remove old files (skip) (#0)

ggml : sync remnants (skip) (#0)

mtgpu: Add MUSA_DOCKER_ARCH in Dockerfiles && update cmake and make (llama/10516)

ggml : add support for dynamic loading of backends (llama/10469)

ggml : sync resolve (skip) (#0)

CUDA: remove DMMV, consolidate F16 mult mat vec (llama/10318)

ggml : build backends as libraries (llama/10256)