chore: some optimization

[iziang]

1. support BGE and Bert embedding
2. Optimize the source module to make it more independent and not rely on kubechat, making it easier to be called externally.
3. Optimize the document QA module by separating the embedding and LLM query processes from the WebSocket consumer, making these two processes more independent and facilitating external calling and testing.
4. Optimize the synchronization module by unifying the initialization of collection creation into the synchronization process, making the code more concise.
5. Support comparing the performance of different embedding models on the same dataset.
6. Add a unique constraint for document names within one collection.
7. Add the fields "peer_type" and "peer_id" to the chat table. Create separate chat records for Feishu one-on-one chats and group chats, enabling storage of Feishu chat records on the KubeChat side.
8. Support exporting Feishu documents to PDF or Word format.
9. Ignore the Feishu document page that contains the string "子页面目录" which is a category and has nothing helpful information
10. Ignore the small document which has only one line and the total size is less than 30.

[iziang]

Ignore the Feishu document page that contains the string "子页面目录" which is a category and has nothing helpful information

# 5.2.1 Python SDK
# 子页面目录

[iziang]

Ignore the small document which has only one line and the total size is less than 30.

# 5.3.1 TODO