Make _emulated_mxfp8_scaled_grouped_mm_2d_2d torch.compile compatible

[lyprince]

• Replace loop-based dequantization in _emulated_mxfp8_scaled_grouped_mm_2d_2d with vectorized ops, removing the @torch.compiler.disable decorator
• The previous implementation used offs.tolist() and a Python loop with data-dependent control flow, which prevented torch.compile from tracing through the emulated MXFP8 wgrad path
• Remove the pytest.skip for EMULATED + use_compile=True in test_mxfp8_grouped_gemm_with_dq_fwd_bwd

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/3906

• 📄 Preview Python docs built from this PR

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

✅ No Failures

As of commit 1fd57f5 with merge base 396d78b ():
💚 Looks good so far! There are no failures yet. 💚

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

[danielvegamyhre]

this is much better, thanks!

[lyprince]

Looks like CI is blocked by PR label check?