vLLM: Revision pinning does not propagate to all artifact load paths

Revision pinning in vLLM does not consistently apply to all artifacts loaded for a model. When an operator supplies --revision or --code-revision, dynamic code, GGUF files, image processors, retrieval side weights, or same-repository subfolder weights/config may still be loaded from an unpinned/default revision.

Operators who pin vLLM model deployments to a reviewed Hugging Face revision for safety review, provenance, rollback, or reproducibility.

Update vLLM to the fixed version (PR #42616) or apply the patch to ensure revision pins propagate to all artifact load paths.