Caching LLM responses is just content addressing

An LLM review costs money and a few seconds of latency. Reviewing the same diff twice should cost...