feat: add vime rollout logp probe

[inaniloquentee]

Summary

• Add an opt-in Vime --custom-generate-function-path shim that preserves Vime native generation and invokes RL-Kernel logp as a minimal rollout-side probe.
• Record structured invocation/fallback evidence under Sample.metadata["rl_kernel"]["vime_logp_probe"].
• Add a mock smoke test that exercises the adapter without full Vime dependencies.
• Document the [WS5] vime native minimal RL baseline #117 fully-async baseline configuration change, expected metadata evidence, and fallback behavior.

Scope

This is a minimal PoC for #120.

It does not replace vLLM sampling/logprob computation, does not change Vime/slime/vLLM source, does not add a production integration, and does not claim train-inference numerical consistency or speedup.

Depends on #117, #118, and #119
Closes #120
Part of #83

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: e9a3f8df-18ee-4901-9ebf-49f8006f5bba

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch feat/ws5-vime-rl-kernel-poc

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[github-actions]

This pull request has been inactive for 30 days. To keep things moving, please update it or let us know if it's still in progress. Otherwise it will be automatically closed in 30 days.