[BugFix][Model] Jamba - Handle aborted requests, Add tests and fix cleanup bug

[mzusman]

Following #4115 , This PR addresses some issues we've found, and adding tests to Jamba:

1. BugFix - Mamba cache slot mapping is now properly cleaned up when finished_requests_ids + current_running_requests > max_mamba_cache_capacity.
2. Added aborted requests to finished_requests_ids inside the scheduler so they will be forwarded to the Jamba's inner state to be properly cleaned up.
3. Added HasInnerState interface to be able to pass SchedulerConfig to the Jamba modeling file,
   SchedulerConfig is used by the Jamba's inner state to determine the max block capacity for the Mamba cache.
4. Added few tests for Jamba

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.
Just a reminder: PRs would not trigger full CI run by default. Instead, it would only trigger fastcheck CI to run, which consists only a small and essential subset of tests to quickly catch errors with the flexibility to run extra individual tests on top (you can do this by unblocking test steps in the Buildkite run).

Full CI run is still required to merge this PR so once the PR is ready to go, please make sure to run it. If you need all test signals in between PR commits, you can trigger full CI as well.

To run full CI, you can do one of these:

• Comment /ready on the PR
• Add ready label to the PR
• Enable auto-merge.

🚀

[mzusman]

/ready

[DarkLight1337]

Thanks for improving the robustness of the code!

[mzusman]

@DarkLight1337 AFAIU pipeline-parallelism-test occasionally fails regardless of this PR, I think it's ready to be merged.

[youkaichao]

AFAIU pipeline-parallelism-test occasionally fails regardless of this PR

any examples for this? we need to investigate.

[DarkLight1337]

AFAIU pipeline-parallelism-test occasionally fails regardless of this PR

any examples for this? we need to investigate.

https://buildkite.com/vllm/ci-aws/builds/4934#0190b6d0-4616-40e5-87cc-b97823315309

I think it's just a connection issue.

[youkaichao]

[2024-07-15T14:59:29Z] (RayWorkerWrapper pid=6923) WARNING 07-15 14:59:29 custom_all_reduce.py:127] Custom allreduce is disabled because your platform lacks GPU P2P capability or P2P test failed. To silence this warning, specify disable_custom_all_reduce=True explicitly.
[2024-07-15T15:08:55Z] ERROR

from the log, it seems the server is stuck somewhere for 8 minutes. difficult to debug.