Update GGUF CUDA kernel code path with MMQ support

[Isotr0py]

What does this PR do?

• Enable MMQ code path with new GGUF CUDA kernel with MMA (standard quant) support.
• Remove MMVQ code path because it's unlikely to be used in diffusion model without decoding stage.

Benchmark Results
GPU: RTX 3090, Model: Flux-1.0-dev + Q4_0 transformer

Native dequantize mulmat

28/28 [00:52<00:00,  1.86s/it]

Kernel dequantize mulmat

28/28 [00:47<00:00,  1.70s/it]

Kernel quantized mulmat (MMQ)

28/28 [00:43<00:00,  1.56s/it]

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline?
• Did you read our philosophy doc (important for complex PRs)?
• Was this discussed/approved via a GitHub issue or the forum? Please add a link to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• Did you write any new necessary tests?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[dxqb]

Native dequantize mulmat

28/28 [00:52<00:00,  1.86s/it]

It might be worth benchmarking any custom kernel against native, but also native compiled by torch.compile, because torch.compile has a major impact on the dequantization code

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.