feat: 工具/技能调用统计 + 用户活动审计 + 标签多选筛选

- 任务1: 用户管理/定时任务管理移入左下角菜单, 从侧边栏删除原入口
- 任务2: 管理员查看用户问答(读 LangGraph checkpoint)与记忆, 新增 admin_users 路由与审计页
- 任务3: 标签筛选改为顶部多选标签栏(OR), agents/skills/scheduled_tasks 三端
- 任务4: 新增 tool_call_metrics 表 + ToolMetricsMiddleware + tool-metrics 路由,
  工具/技能调用统计并入大模型统计页, 含错误分类
- 技能识别: lead agent 经 read_file 加载 /mnt/skills/<cat>/<name>/SKILL.md 调用技能,
  中间件扫描工具参数里的 skills/{public,custom}/{name}/ 路径还原 skill_name
- 单元测试 tests/test_tool_metrics_middleware.py (18 项全过)
- 附: SQLite->MySQL 切换后清理脚本

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

frontend-web/src/core/artifacts/loader.ts

@@ -9,16 +9,18 @@ export async function loadArtifactContent({
   filepath,
   threadId,
   isMock,
+  publicShareCode,
 }: {
   filepath: string;
   threadId: string;
   isMock?: boolean;
+  publicShareCode?: string;
 }) {
   let enhancedFilepath = filepath;
   if (filepath.endsWith(".skill")) {
     enhancedFilepath = filepath + "/SKILL.md";
   }
-  const url = urlOfArtifact({ filepath: enhancedFilepath, threadId, isMock });
+  const url = urlOfArtifact({ filepath: enhancedFilepath, threadId, isMock, publicShareCode });
   const response = await apiFetch(url);
   const text = await response.text();
   return { content: text, url };