Update CI for HF Optimum models (#10820)

### Summary
Updated pinned `optimum-executorch`. Use `optimum-cli` to generate pte
files with xnnpack+custom_sdpa+8da4w. Getting the CI ready to be
profiling those ptes using `executor_runner`.

### Test plan
CI will export native HF models via `optimum-cli` and validate those
models via python API `from_pretrained`.

Co-authored-by: Guang Yang <guangyang@fb.com>