fix: harden L3 runtime continuity and tool execution

Align the L3 graph, agent service, and sync tool shims on one canonical continuity contract so clarification resumes and persisted snapshots behave consistently. Add targeted regressions and hardening notes covering system-message coalescing, async bridge usage, and continuity rehydration.

backend/app/agents/tools/search.py

@@ -5,25 +5,16 @@ Agent 工具集 - 知识库 & 图谱相关
 由于 LangChain 工具系统是同步的，内部用 run_in_executor 处理 async 逻辑。
 """
 
-from concurrent.futures import ThreadPoolExecutor
-import asyncio
-
 from langchain_core.tools import tool
 
 from app.agents.context import get_current_user
+from app.agents.tools.async_bridge import run_async
 from app.database import async_session
 
-_executor = ThreadPoolExecutor(max_workers=4)
-
 
 def _run_async(coro, timeout: int = 30):
     """在同步上下文中运行 async 代码"""
-    try:
-        loop = asyncio.get_running_loop()
-        future = loop.run_in_executor(_executor, lambda: asyncio.run(coro))
-        return future.result(timeout=timeout)
-    except RuntimeError:
-        return asyncio.run(coro)
+    return run_async(coro, timeout=timeout)
 
 
 @tool