Source transform to use static attention

[sxu]

Summary: Introduce a source transform to be more aligned with other transforms we run, also makes it less error prone (e.g. HF RoPE transformation needs to happen before turning linears into conv2ds).

Differential Revision: D84769599

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/15176

• 📄 Preview Python docs built from this PR

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

❌ 2 New Failures, 1 Unrelated Failure

As of commit 1bc10dd with merge base e1d9fd2 ():

NEW FAILURES - The following jobs have failed:

• pull / test-samsung-models-linux / linux-job (gh)
  RuntimeError: Command docker exec -t bda1422eec2b64b0c1cf28c437190cd2c0073c68d0b3f1073ca30d0483595f44 /exec failed with exit code 1
• Test CUDA Builds / export-voxtral-cuda-artifact / linux-job (gh)
  RuntimeError: Command docker exec -t 7cd14b01c41670d17e6fcfdee6ab3e3cf8515f2aeb97b5369c55cc32f75b7fc8 /exec failed with exit code 2

FLAKY - The following job failed but was likely due to flakiness present on trunk:

• pull / unittest-arm-backend-with-no-fvp (test_pytest_models) / linux-job (gh) (detected as infra flaky with no log or failing log classifier)

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

[meta-codesync]

@sxu has exported this pull request. If you are a Meta employee, you can view the originating Diff in D84769599.