Change linux.gcp.a100 to linux.gcp.a100.large (#95913)
To avoid making workloads like https://github.com/pytorch/pytorch/blob/master/.github/workflows/inductor.yml#L52 queue for a long time.
For example, in the past the runners were all used to run https://github.com/pytorch/pytorch/actions/workflows/inductor-perf-test-nightly.yml and perf smoke test jobs in https://github.com/pytorch/pytorch/actions/workflows/inductor.yml did not get runners.
<img width="614" alt="image" src="https://user-images.githubusercontent.com/109318740/222570066-7aec611d-0feb-42cb-8b1b-d93bd36f4d17.png">
This PR makes sure the queue only happens to long-running workloads and we have a plan to address it with more runners with basic auto-scaling feature enabled.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/95913
Approved by: https://github.com/malfet, https://github.com/huydhn
