LLM 0.37.2 fixes API key leak to third-party OpenAI-compatible endpoints

Simon Willison released LLM 0.37.2 to fix a security issue reported by Daniel de Araujo in GHSA-v9h7-jj4g-f862 and GHSA-r639-rr7r-x72w. The bug could send a configured OpenAI API key to third-party OpenAI-compatible endpoints via extra-openai-models.yaml if the named api_key_name was never stored, or if a responses: true model ran with -o chat_completions 1.