perf(hub-ai): reduce CPU load during trading coach chat

Cache chat context, parallelize exchange fetches, skip fund history writes, defer rolling summary to a background thread, and cache markdown rendering on the client.

Co-authored-by: Cursor <cursoragent@cursor.com>

manual_trading_hub/hub_ai/routes.py

@@ -1,6 +1,7 @@
 """中控 AI FastAPI 路由。"""
 from __future__ import annotations
 
+import asyncio
 from typing import Callable
 
 from fastapi import APIRouter, Body, File, Form, HTTPException, UploadFile
@@ -153,7 +154,8 @@ def create_hub_ai_router(*, load_all_exchanges: Callable[[], list]) -> APIRouter
                     "data": data,
                 }
             )
-        result = send_chat_message(
+        result = await asyncio.to_thread(
+            send_chat_message,
             exchanges,
             message,
             trading_day=_day(trading_day) if trading_day.strip() else None,