Fix static typing issues in v1/attention

[hmellor]

• AttentionMetadata and AttentionMetadataBuilder are not inherited in V1, so using them as generic return types doesn't work
• Instead we add generic typing to AttentionBackend so that the different backends can tell AttentionBackend exactly which return types to expect
• All other typing issues in the v1/attention directory have ben fixed
  • vllm_flash_attn doesn't exist in a fresh clone (i.e. pre-commit CI)
• v1/attention has been added to the list of directories that is more strict in CI

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can either: Add ready label to the PR or enable auto-merge.

🚀

[hmellor]

Closing this as stale. May revisit once ty is ready