[MLX] Add Q5_K quantization support (#20617) #14373

Job	Run time
Get changed files / get-changed-files	4s
CI run decision / decide	32s
test-executorch-cuda-build-${{ matrix.cuda-version }}	0s
unittest-cuda	0s
test-models-cuda	0s
unittest-cuda-runtime	0s
test-cuda-pybind	0s
export-model-cuda-artifact	0s
check-all-cuda-builds	0s
test-model-cuda-e2e	-1s
	35s

Provide feedback