fix: acquire session lock in background and cron wake paths

[T52T52]

问题
后台任务唤醒路径（wakemainagentforbackgroundresult）和定时任务唤醒路径（wokemain_agent）都绕过了 sessionlockmanager，导致同一会话的 agent 工作流与用户消息处理并发执行，造成：

• LLM 上下文过期（第二个请求看不到第一个请求的内容）
• 对话历史丢失（persistagenthistory 中并发读改写互相覆盖）

复现步骤
1.让llm同时调用两个或以上子agent
2. 多个子 agent 同时返回结果，主 agent 的唤醒路径并发触发
3.历史被覆盖，llm只有最后一个子agent回复的上下文

根因
用户消息处理路径（agentsubstages/internal.py）将整个 agent 流程包在 sessionlockmanager.acquirelock(unifiedmsg_origin) 中，但两条唤醒路径直接调用 buildmainagent，没有获取会话锁。

修复
将两条唤醒路径的「读取-构建-运行-持久化」流程包进 per-session 锁中，与用户消息路径保持一致。

测试

• tests/testbackgroundwake_lock.py
• tests/testcronwake_lock.py
• tests/testpersistrace.py

全部 4 个测试通过。修复前 testbackgroundwake_lock 和 testcronwake_lock 失败（acquirelock 调用 0 次），testpersistrace 丢失 3/4 并发结果。

Summary by Sourcery

Guard background and cron agent wake workflows with per-session locking to prevent concurrent access to conversation history.

Bug Fixes:

• Ensure background-task wake path acquires the session lock before reading, running the main agent, and persisting history.
• Ensure cron wake path acquires the session lock before reading, running the main agent, and persisting history.
• Prevent loss of conversation context and assistant results caused by racing writes to persisted agent history.

Tests:

• Add regression tests confirming background wake acquires the session lock and no longer runs agents without session serialization.
• Add regression tests confirming cron wake acquires the session lock and respects per-session serialization.
• Add a concurrency regression test for persist_agent_history to verify that multiple concurrent writes preserve all results in the session history.

[sourcery-ai]

Hey - I've reviewed your changes and they look great!

---

Sourcery is free for open source - if you like our reviews please consider sharing them ✨

• X
• Mastodon
• LinkedIn
• Facebook

_{Help me be more useful! Please click 👍 or 👎 on each comment and I'll use the feedback to improve your reviews.}