[torchfuzz] add sdpa operator

[bobrenjc93]

Stack from ghstack (oldest at bottom):

• [torchfuzz] split, chunk, stack, cat, expand, gather, cumsum, clamp, index_select, split #166221
• [torchfuzz] add mhaf operator #166190
• -> [torchfuzz] add sdpa operator #166189
• [torchfuzz] make pointwise subclasses defined torch_op_name #166220
• [torchfuzz] fix group norm operator #166188
• [torchfuzz] check in more ignore regexes #166187

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/166189

• 📄 Preview Python docs built from this PR
• 📄 Preview C++ docs built from this PR
• ❓ Need help or want to give feedback on the CI? Visit the bot commands wiki

Note: Links to docs will display an error until the docs builds have been completed.

✅ You can merge normally! (2 Unrelated Failures)

As of commit e6055a7 with merge base 78bcfcf ():

FLAKY - The following jobs failed but were likely due to flakiness present on trunk:

• pull / linux-jammy-py3.10-gcc11 / test (distributed, 1, 2, linux.2xlarge) (gh) (similar failure)
  test/distributed/pipelining/test_microbatch.py::MicrobatchTestsCPU::test_split_block_mask_batch_size_one_cpu
• trunk / linux-jammy-cuda12.8-py3.10-gcc11 / test (distributed, 1, 3, lf.linux.g4dn.12xlarge.nvidia.gpu) (gh) (similar failure)
  test/distributed/pipelining/test_microbatch.py::MicrobatchTestsCUDA::test_split_block_mask_batch_size_one_cuda

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[pytorchmergebot]

Starting merge as part of PR stack under #166221