[Core] Async_output_proc: Add virtual engine support (towards pipeline parallel)

[alexm-neuralmagic]

This a small PR that adds virtual engine support to the asynchronous post-processor, which is necessary to allow pipeline parallel.

[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 run fastcheck CI which consists a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of default ones by unblocking the steps in your fast-check build on Buildkite UI.

Once the PR is approved and ready to go, please make sure to run full CI as it is required to merge (or just use auto-merge).

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

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

🚀

[alexm-neuralmagic]

/ready

[alexm-neuralmagic]

@megha95 @WoosukKwon @comaniac @robertgshaw2-neuralmagic @SolitaryThinker

[njhill]

I feel this name is a bit generic? Maybe something like request_output_callback is better? i.e. to make it clearer what this is calling you back for/with

[alexm-neuralmagic]

I would prefer to have it async_callback because it is not only output_proc anymore in the async + multi-step PR. The callback starts to get more and more functionality inside.

[comaniac]

LGTM. Leave to @njhill

[WoosukKwon]

@alexm-neuralmagic Can you please address @njhill's comments?

[alexm-neuralmagic]

Addressed comments, @njhill thanks for the good suggestion with partial funcs!

[njhill]

Thanks @alexm-neuralmagic