Run a bounded review-and-fix loop on a pull request, fixing only findings that carry a failure scenario. Use when the user asks for a PR review loop, asks to keep reviewing and fixing until clean, or right after a pull request has been created.
