XNNPACK workflow don't stop on one run error (#1077)

Summary:
Instead, need to wait for all options (quantization/delegation) and report error in the end.

See https://github.com/pytorch/executorch/actions/runs/6631014744/job/18013650086?pr=1077, now we run all models, and see useful info like `Portable q8 ok, Delegation fp32 error, Delegation q8 error`

Pull Request resolved: https://github.com/pytorch/executorch/pull/1077

Reviewed By: huydhn

Differential Revision: D50610446

Pulled By: kirklandsign

fbshipit-source-id: 1985c519026b8a28ad5512f993e65b88254694fb