Basics of Threading and Improvements to Assistants

We originally had this proposed resolution here. However its explored more in the child issue.

• Can we make it so that it chooses when to use RAG, rather than always doing a RAG search. This could be achieved via a simplechat model to save money (Such as ChatGpt 4o mini) - This is explored in this child issue: https://www.drupal.org/project/ai/issues/3470263 🌱 Handle threading and history with Assistants API and RAG Active
• Implement "SimpleChat" - default -
• Chatbot should be able to receive files. (Where are those files stored if we don't store history?)
• If we decide a simple model decides if its RAG or not check the flag for JSON
• Add the ability for a "Pre Search" prompt (Also what should it tell the next LLM - Provide a default answer) and a "Post search Prompt". If its empty don't use it. (DON'T DO THIS, BUT NEE TO MAKE SURE THIS USE CASE IS SORTED