feat(server): add Notebook Q&A RAG Chat API endpoint (POST /api/extended/chat)

[allenporter]

Feature Request: Notebook Q&A / RAG Chat API (POST /api/extended/chat)

Summary
Add a server endpoint POST /api/extended/chat to allow conversational multi-turn Q&A over transcribed notebooks.

Details
While the backend currently provides vector semantic search (/api/extended/search), returning raw matching page chunks, it lacks a high-level conversational synthesis endpoint that takes a user query, retrieves relevant embeddings, and calls Gemini to synthesize an answer with page citations.

Proposed API Specification

POST /api/extended/chat
Request:
{
  "prompt": "What were the key architectural decisions made in Sprint Planning?",
  "fileId": 739205171027902477,
  "top_n": 5
}

Response:
{
  "success": true,
  "answer": "The team decided to migrate SQLite database schemas using Alembic...",
  "citations": [
    { "fileId": 739205171027902477, "fileName": "Architecture.note", "pageIndex": 0 }
  ]
}

MCP Considerations
Consider whether this endpoint can reuse or complement the MCP server tools (supernote/server/mcp/server.py).