feat: Mermaid is now directly output by LLM

[Lilia-Chen]

Context

A part of the issue https://github.com/sheepbox8646/ChatTutor/issues/21

Work in progress

Still can't correctly parse the block head...

[Lilia-Chen]

Hi @sheepbox8646, I need your review on this due to the important change in the code.
Currently, LLM can directly output Mermaid code, and the agent/service layer can automatically parse it as a Mermaid page and render it on the front end. The created page will be pushed into pages, so LLM still know them in the message/memory.
As a next step, the current code can be extended to parse note pages and code blocks.