Describe the feature

Align with vllm, retrieve X-Request-Id from request headers as request_id or generating a new uuid as request_id only when X-Request-Id is not present in the request headers.
In vllm/vllm/entrypoints/openai/serving_completion.py

    @staticmethod
    def _base_request_id(raw_request: Optional[Request],
                         default: Optional[str] = None) -> Optional[str]:
        """Pulls the request id to use from a header, if provided"""
        default = default or random_uuid()
        if raw_request is None:
            return default

        return raw_request.headers.get("X-Request-Id", default)

Why do you need this feature?

1. Improve request tracing in serving systems: Allow upstream services to pass request_id to the router, maintaining request_id consistency throughout the service chain
2. Simplify troubleshooting: External systems can provide their own request_id, making cross-system tracing and debugging easier.

Additional context

No response