Optimize performance and scalability, and LLM conversation summaries

This project aims to optimize the performance and scalability of many code paths in Zulip, with a focus on performance sensitive code paths like Zulip major endpoints which have request volume and cost that makes them important. The project also aims to design and implement a new feature of “LLM conversation summary”, and includes fixing issues in other areas.