pub fn make_inference_request_span<B>(req: &Request<B>) -> SpanExpand description
Create a span for inference request endpoints (completions, chat, embeddings, etc.).
Uses target: "request_span" which is always allowed through the DYN_LOG filter
(via request_span=trace directive in filters()). This ensures request context
(request_id, model, trace_id) is always available on log events.