[metrics] Add proton profiling

[xuzhao9]

Add Proton profiler support. The profiler will instrument into the run() function to instrument the benchmark process.

Test plan:

$ python run.py --op softmax --num-inputs 3 --metrics proton
Entering scope x_val_[4096, 256]
Exiting scope x_val_[4096, 256]
Entering scope x_val_[4096, 384]
Exiting scope x_val_[4096, 384]
Entering scope x_val_[4096, 512]
Exiting scope x_val_[4096, 512]
      x_val    naive_softmax-proton    triton_softmax-proton
-----------  ----------------------  -----------------------
[4096, 256]
[4096, 384]
[4096, 512]

$ proton-viewer  proton.hatchet -m time

[facebook-github-bot]

@xuzhao9 has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator.

[facebook-github-bot]

@xuzhao9 has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator.

[adamomainz]

Excellent! Great job getting this working

[facebook-github-bot]

@xuzhao9 merged this pull request in 87dffcc.