[code_review] LLM as judge to validate comments

Wrong comments can make people spend a lot of time for nothing, greatly reducing the trust in the tool. The LLM as judge could have access to the same code search tools that comment generation has, plus docs.