feat: enhance RAG ingestion, citations, imports, and knowledge/memory…

[WeiliangShen]

This change upgrades the repository’s RAG pipeline with multi-format document ingestion, structure-aware chunking, query-aware retrieval, stable citations, incremental knowledge import indexing, and web console support for chunk inspection, knowledge creation/deletion, and memory deletion.
Please let me know if this should target another branch.

[zhayujie]

Thanks for this nice work, Multi-format ingestion + smarter chunking is something we've been wanting to add.

This is a fairly large change and has conflicts with master, so I'd like to land it on a feature branch first and test things out before promoting to master.

I've created a dev-rag-enhance branch for this. Could you retarget this PR to dev-rag-enhance and resolve the conflicts? I'll review and merge it there, then do some testing and adjustments on my side (a few things like the background watcher thread and asyncio.run() usage I want to tweak) before merging into master.