Reduce macOS CI jobs and add more Arm64 jobs. (#9228)

Depends on https://github.com/pytorch/executorch/pull/9227/ and
https://github.com/pytorch/executorch/pull/9207

Here's the net result after a sequence of PRs:

- Reduce trunk test-model-macos jobs (from 38 down to 15)
- Add arm64 trunk test-model jobs (from 0 up to 31)
- Add arm64 pull test-model-jobs jobs (+4)