[Attention] Refactor CUDA attention backend selection logic

[MatthewBonanni]

Purpose

CudaPlatformBase.get_attention_backend_cls has gotten complex and messy over time. This PR cleans up the logic (without changing the behavior) and standardizes the interface.

Test Plan

Test Result

---

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.
• (Optional) Release notes update. If your change is user facing, please update the release notes draft in the Google Doc.

[mergify]

This pull request has merge conflicts that must be resolved before it can be
merged. Please rebase the PR, @MatthewBonanni.

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

[mergify]

This pull request has merge conflicts that must be resolved before it can be
merged. Please rebase the PR, @MatthewBonanni.

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

[LucasWilkinson]

left a few comments; we should figure out who the owner of the plugin mechanism is and figure out how to notify downstream HW plugins since I think this will affect them pretty dramatically

[MatthewBonanni]

@LucasWilkinson thanks for your review! I've already notified @ILikeIneine but I'm not sure if there's anyone else we should reach out to?

[ILikeIneine]

@MatthewBonanni Hi, would this refactor be able to merge into v0.11.1?

[MatthewBonanni]

@ILikeIneine we were planning on waiting until after v0.11.1, we don't want to risk further delaying the release and because it changes the platform interface, it might be better to be part of v0.12

[LucasWilkinson]

@MatthewBonanni how hard would it be to keep backwards compatibility between _Backend and AttentionBackendEnum for a version with a warning?

[LucasWilkinson]

With #26487 potential in the pipe what do we think about having a get_mla_attn_backend_cls instead of is_mla? @Yikun ?

[MatthewBonanni]

@MatthewBonanni how hard would it be to keep backwards compatibility between _Backend and AttentionBackendEnum for a version with a warning?

@LucasWilkinson done in d0f4698

[NickLucche]

Discussed offline thanks for the work @MatthewBonanni !