[V1] [Hybrid] Add new test to verify that hybrid views into KVCacheTensor are compatible

[tdoublep]

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.

Purpose

When running hybrid (attention/mamba) models in V1, the different types of layers share the same KVCacheTensor, by having different views into it. When implementing the support, care had to be taken to ensure that these two views are "compatible" in the sense that writing to a block in the mamba view will not corrupt the data in a different block in the attention view. I have found mistakes here are a bit tricky to catch since they don't necessarily show up until you start filling up all of the blocks. This PR adds a test to explicitly verify the compatibility of these views.

I would like to have this test in place before we verify the solution in #21123 to the large block size problem. It will also be helpful to have the test in place when we add the additional support for attention backends with (2, num_blocks) layout.

cc @heheda12345 @tlrmchlsmth

Test Plan

See diff

Test Result

Passing locally

(Optional) Documentation Update

[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.

🚀

[gemini-code-assist]

Code Review

This pull request adds a new test to verify the compatibility of hybrid attention/mamba views into the KVCacheTensor. To ensure full compatibility, the test should also verify that the mamba blocks retain their expected values after the attention blocks are written.

[heheda12345]

Thank you very much.