[CUDA] Add Triton Q4_K-to-FP8 prefill linear

[Gasoonjia]

Stack dependency

Depends on #22190.

Summary

Add an SM90+ pure-Triton Q4_K-to-FP8 prefill linear path.

• Dequantize Q4_K weights directly to FP8 and feed an FP8 matrix multiplication with BF16 output.
• Retain the existing low-token decode kernel and BF16 prefill fallback for unsupported devices or shapes.
• Add float8 dtype support to the slim AOT runtime shims.
• Select the CUDA-target FP8 custom op during export even when export example tensors reside on CPU.

This change remains Triton-only and does not introduce CUDA or CuteDSL kernels.

Operator-level performance

RTX 5090, representative Muse MLP projection (N=39,936, K=6,656), Q4_K dequantization plus linear:

Prefill rows M	BF16 dequant + linear	Q4_K-to-FP8 path	Speedup
128	5.033 ms	0.507 ms	9.93x
512	5.818 ms	0.945 ms	6.16x
2,048	9.182 ms	2.688 ms	3.42x

Test plan

• lintrunner init && lintrunner -a (clean).
• Int4 dispatch tests: 22 passed.
• Verified that SM90+ export captures triton.q4k_fp8_linear.
• Final PTE profile contains 208 FP8 matmul and 208 Q4_K-to-FP8 dequant kernel calls.
• git diff --check.

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/22192

• 📄 Preview Python docs built from this PR

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

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