Tighten TQ4 SDPA input validation

[fallintoplace]

Summary

Tightens the Python wrapper validation for the TQ4 SDPA kernel so malformed inputs fail before shape unpacking, indexing, reshaping, or launching Triton.

Changes include:

• validate ranks first for query, packed K/V, norms, centroids, rotation, mask, and kv_len
• validate all tensors are on the same CUDA device as query
• validate dtypes before layout-sensitive shape checks
• validate packed K/V shape symmetry, norm layout, centroids shape, rotation shape, mask shape, and kv_len scalar shape
• add malformed-input coverage for 3D query, 3D mask, bad rotation, bad centroids, CPU k_norms, and v_packed shape mismatch

Test Plan

• python3 -m py_compile backends/cuda/triton/kernels/tq4_sdpa.py backends/cuda/tests/test_tq4_sdpa.py
• git diff --check -- backends/cuda/triton/kernels/tq4_sdpa.py backends/cuda/tests/test_tq4_sdpa.py

I could not run the targeted CUDA unittest locally because the available Python environment is missing numpy and torch.

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/20410

• 📄 Preview Python docs built from this PR

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

❗ 1 Active SEVs

There are 1 currently active SEVs. If your PR is affected, please view them below:

• [ROCm] MI350 CI jobs will have longer queue times due to CI migration

⚠️ 13 Awaiting Approval

As of commit d906314 with merge base 55c54c7 ():

AWAITING APPROVAL - The following workflows need approval before CI can run:

• Build documentation (gh)
• Build Presets (gh)
• Lint (gh)
• periodic (gh)
• pull (gh)
• Test ARM Backend (gh)
• Test CoreML Backend (gh)
• Test CUDA Builds (gh)
• Test CUDA Windows Export and E2E (gh)
• Test QNN Backend (gh)
• Test Vulkan Backend (gh)
• Test WebGPU Backend (gh)
• Test XNNPACK Backend (gh)

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

[linux-foundation-easycla]

The committers listed above are authorized under a signed CLA.

• ✅ login: fallintoplace / name: Minh Vu (d906314)

[fallintoplace]

@pytorchbot label "release notes: none"