test: enable bmm_mxfp8 cutlass backend coverage on SM12x (#3183)

## 📌 Description

`test_bmm_mxfp8.py` only parametrizes `backend=["cudnn"]` and skips
`cap[0] in [11, 12]` outright, so the cutlass backend (which is the one
actually supported on SM12x — see `gemm_base.py:7570`) was never
exercised. The cudnn backend is supported on `[100, 103]` only, so
simply removing the skip would have failed on SM12x.

This adds `cutlass` to the backend parametrize and replaces the
unconditional skip with per-backend CC gating that mirrors the
`@supported_compute_capability` decorators:

  - `cudnn` backend: `cap[0] != 10` → skip
  - `cutlass` backend: `cap[0] != 12` → skip

  After the change:
  - SM10x: runs cudnn (unchanged), skips cutlass
  - SM12x: skips cudnn, runs cutlass (newly covered)
  - SM110: skipped on both (separate gap, not addressed here)

  Verified on RTX Pro 6000 (sm_120, CUDA 12.9):
  - cutlass backend: **288 passed, 0 failed**
  - cudnn backend: 288 skipped (as expected on sm_120)

  ## 🔍 Related Issues

  Refs #3170 (Action Item 7 / T3).

  ## 🚀 Pull Request Checklist

  ### ✅ Pre-commit Checks

- [x] I have installed `pre-commit` by running `pip install pre-commit`
(or used your preferred method).
  - [x] I have installed the hooks with `pre-commit install`.
- [x] I have run the hooks manually with `pre-commit run --all-files`
and fixed any reported issues.

  ## 🧪 Tests

  - [ ] Tests have been added or updated as needed.
  - [x] All tests are passing (`unittest`, etc.).

  ## Reviewer Notes



<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit

* **Tests**
* Expanded test coverage to include an additional GPU computation
backend, validating batch matrix multiplication across backends.
* Strengthened GPU compatibility checks with backend-specific
requirements and added conditions for a particular memory layout,
ensuring tests run only on supported hardware/configurations.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->

---------

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
Co-authored-by: Ka-Hyun Nam <knam@nvidia.com>