Add new torchao experimental a8wxdq for fast ARM CPU quantization #2765

Job	Run time
gather-models-cpu	9s
compile-gguf (macos-14)	5m 15s
test-gguf-util (macos-14)	5m 44s
gather-models-gpu	13s
test-tinystories-executorch (16-core-ubuntu)	7m 3s
test-tinystories-executorch (macos-14-xlarge)	6m 44s
runner-et (16-core-ubuntu)	8m 34s
runner-aoti (16-core-ubuntu)	6m 17s
runner-et (macos-14-xlarge)	8m 17s
runner-aoti (macos-14-xlarge)	3m 21s
test-torchao-experimental (macos-14-xlarge)	1m 34s
torchchat-command-load-test (macos-14)	2m 28s
test-mps / macos-job	1m 59s
test-mps-dtype / macos-job	2m 24s
test-build-runner-et-android / linux-job	13m 30s
test-cpu-eval-sanity-check (x86_64, stories15M)	10m 26s
test-cpu-eval-sanity-check (aarch64, stories15M)	10m 7s
test-cpu-compile (x86_64, stories15M)	17m 0s
test-cpu-eval-sanity-check-float16 (x86_64, stories15M)	30m 23s
test-cpu-compile (aarch64, stories15M)	9m 42s
test-cpu-aoti (x86_64, stories15M)	17m 22s
test-cpu-eval-sanity-check-float16 (aarch64, stories15M)	6m 59s
test-cpu-aoti (aarch64, stories15M)	9m 38s
test-cpu-eval-sanity-check-float32 (x86_64, stories15M)	39m 0s
test-cpu-eval-sanity-check-float32 (aarch64, stories15M)	9m 1s
test-gpu-aoti-float16 (cuda, stories15M) / linux-job	12m 33s
test-gpu-compile (cuda, stories15M) / linux-job	23m 44s
test-gpu-aoti-bfloat16 (cuda, stories15M) / linux-job	13m 30s
test-gpu-aoti-float32 (cuda, stories15M) / linux-job	12m 28s
test-gpu-eval-sanity-check (cuda, stories15M) / linux-job	12m 15s
	5h 7m 40s

Provide feedback