[ROCM] Add MI350 support for MXFP8 colwise quantization.

[xiaobochen-amd]

Summary

Add ROCm MI350 (gfx950) support for MXFP8 quantization kernel.

Changes

• Implement mxfp8_quantize for ROCm in mxfp8_extension.cpp and mxfp8_rocm.hip
• Support colwise quantization with column-major output layout (matching CUDA API)
• Support both FLOOR and RCEIL scaling modes
• Add MI350 to test conditions in test_kernels.py

Testing

• Validated against CUDA reference implementation on MI350
• All test_cuda_mx_dim1_numerics tests pass for FLOOR and RCEIL modes

docker:  rocm/primus:v25.10

torch==2.11.0.dev20251221+rocm7.1

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/3544

• 📄 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:

• B200 runners are down due to network issues

❌ 1 New Failure

As of commit 2636ce6 with merge base 57432bd ():

NEW FAILURE - The following job has failed:

• PR Label Check / Check PR Labels (gh)
  Process completed with exit code 1.

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