Cortex-M: quantize aten.matmul by rewriting it to bmm before annotation

[rascani]

Summary

The cortex_m quantizer annotates only aten.bmm.default, but torch.matmul / the @ operator is captured as aten.matmul.default and does not decompose to bmm until to_edge -- after quantization runs -- so matmuls (for example every attention score and context product) never received qparams and stayed in fp32 on portable kernels.

Add a pre-annotation MatmulToBmmPass that rewrites aten.matmul to aten.bmm (rank-3 directly; higher ranks by folding the leading batch dims to 3D and reshaping back; rank-2 and broadcasting matmuls left unchanged), so the existing CortexMBmmCheck annotation and quantized_batch_matmul lowering handle it. No new kernel is needed -- the batch-matmul kernel already exists; this is purely an annotation-timing fix.

Test plan

Verified test-first (RED->GREEN) with rank-3 and rank-4 dialect tests and no regression on the existing bmm suite; confirmed it flips all 15 SAM mask-decoder attention matmuls from fp32 to quantizable.

Authored with Claude Code.

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/21028

• 📄 Preview Python docs built from this PR

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

❌ 1 Unclassified Failure

As of commit 21bdfd3 with merge base 7013c8d ():

UNCLASSIFIED FAILURE - DrCI could not classify the following job because the workflow did not run on the merge base. The failure may be pre-existing on trunk or introduced by this PR:

• Test QNN Windows MSVC build / build-qnn-windows-msvc / windows-job (gh) (this job did not run on the merge base, so DrCI cannot tell whether the failure is pre-existing)
  Process completed with exit code 1.

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

[github-actions]

This PR needs a release notes: label

If your change should be included in the release notes (i.e. would users of this library care about this change?), please use a label starting with release notes:. This helps us keep track and include your important work in the next release notes.

To add a label, you can comment to pytorchbot, for example
@pytorchbot label "release notes: none"

For more information, see
https://github.com/pytorch/pytorch/wiki/PyTorch-AutoLabel-Bot#why-categorize-for-release-notes-and-how-does-it-work.