feat: §17 ToolchainHandler 强约束实现（Step 1-4）

Step 1: 基础设施
- prompt_composer.py: PromptContext 新增 action_type + action_steps 字段
- spawner.py: handler 路径提取 action_type/action_steps 传入 PromptContext
- db.py: comments CHECK 约束加入 action_report

Step 2: ToolchainHandler 强化
- ToolchainContextSection: 加 steps 渲染 + action_hint（按 action_type）
- ToolchainApiSection: 改为 action_report 提交指引 + Gitea 协作指引
- ToolchainConstraintsSection: 5 条强约束 + Red Flags 防self-rationalization
- verify_completion: action_report → output → comment 三层 fallback
  - review_merged 始终通过（纯通知）
  - infrastructure_failure 始终通过（防递归）
  - 修复 LENGTH(content) → LENGTH(body) bug
- on_failure 三分路: 业务→Gitea PR comment / 系统→Gitea Issue / 基础设施→toolchain task

Step 3: toolchain_routes 改造
- 新增 _toolchain_db_path() + _send_toolchain_task()
- 所有 8 个 handler 改为 _send_toolchain_task
- _send_mail 保留但不再被 toolchain handler 调用
- _send_deploy_failure_mail → _send_deploy_failure_task

Step 4: 测试
- 29 个单元测试全部通过
- 全量 456 passed, 3 skipped, 0 failures

src/daemon/spawner.py

@@ -286,10 +286,15 @@ class AgentSpawner:
             # 从 must_haves 解析 mail 元数据（from / performative）
             from_agent = ""
             mail_type = ""
+            action_type = ""
+            action_steps = []
             try:
                 meta = json.loads(must_haves) if must_haves else {}
                 from_agent = meta.get("from", "")
                 mail_type = meta.get("performative", meta.get("type", ""))
+                # toolchain 字段提取
+                action_type = meta.get("action_type", "")
+                action_steps = meta.get("steps", [])
             except Exception:
                 pass
             ctx = PromptContext(
@@ -298,6 +303,7 @@ class AgentSpawner:
                 agent_id=agent_id, role=spawn_type,
                 spawn_type=spawn_type,
                 from_agent=from_agent, mail_type=mail_type,
+                action_type=action_type, action_steps=action_steps,
             )
             return handler.build_prompt(ctx)