Add context to bot conversations

[Agent-E11]

I tried to do this the first time I added llm support, but I ran into the issue that you cannot easily fetch a chain of replies from a message. This probably means that we will have to keep our own cache of contexts, and possibly associate each new message to an existing one, if it is replying to a message that was "in the conversation". Or, we could just store a universal context for the day, and remove it every midnight or something (this might be pretty resource intensive though)

[Agent-E11]

This is probably a good use of an sqlite database

[Agent-E11]

Regarding storage format. A way to distinguish between users could be something like this (in the list of messages):

{
    { "role": "user", "content": "(agent_e11 <@0123456789>) Hello, this is my message." }, // (user_name <@user_id>)
    { "role": "assistant", "content": "Hello, this is my message." } // No markup
}

And then add something like this to the system prompt:

The user messages are prefixed with (user_name <@user_id>),
where the user's name is user_name and the user's id is user_id.