graph : make FA compatible with MLA + add initial Metal kernels (ggml…

…-org#12953)

* graph : make mla compatible with FA

* metal : add exp FA kernels for DeepSeek models

ggml-ci

* llama : minor naming updates

ggml-ci

* ggml : disable FA for DS head sizes

* tests : add FA tests for MLA shapes

ggml-ci

CANN: Add 310P operator support check (ggml-org#12962)

Add performance print for gemma3 in example (ggml-org#12929)