ci: prototype vllm-vcr backend for faster Responses integration tests

[leseb]

Summary

Prototype a vllm-vcr-backed CI lane to measure the speedup from
eliminating the ~8 minute vLLM CPU model startup penalty. The
branch-scoped workflow runs the vLLM frontend with vllm-vcr random
tokens instead of real inference. The test file gains a
VLLM_TEST_BACKEND=vcr-random mode that runs 3 protocol-safe tests
(smoke, store/retrieve, streaming) and skips 7 that need real model
inference or captured tool-call traces.

Related issue

N/A — exploratory prototype to evaluate CI runtime savings.

Validation

• make lint
• make build
• CI prototype workflow (will run on push to this branch)

Checklist

• I reviewed every changed line and can explain the change.
• New capabilities include an example config and functional example test.
• User-facing behavior and generated documentation are updated.
• Performance-sensitive changes include appropriate benchmark or load-test evidence.
• Commits are signed and include a Signed-off-by trailer.

Breaking changes

None — branch-scoped prototype workflow, no changes to existing CI or test behavior.