Runtime tasks: refresh last_run_at when alarm/inventory sync starts

Avoid stale last_run_at when sync_alarms_current runs longer than interval; clarifying tooltip on UME page.

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

netx_api/main.py

@@ -449,6 +449,7 @@ def on_startup() -> None:
                         _set_runtime_task(
                             "alarms_current_auto_sync",
                             status="running",
+                            last_run_at=datetime.now(timezone.utc),
                             last_error="正在拉取 UME 当前告警…",
                         )
                         db = SessionLocal()
@@ -491,6 +492,7 @@ def on_startup() -> None:
                         _set_runtime_task(
                             "inventory_auto_sync",
                             status="running",
+                            last_run_at=datetime.now(timezone.utc),
                             last_error="正在拉取 UME 网元清单…",
                         )
                         db = SessionLocal()

web/src/pages/UmePage.tsx

@@ -296,7 +296,9 @@ export function UmePage({ toastOk, toastError }: UmePageProps) {
               <th>task</th>
               <th title="相邻两次循环 sleep 的配置间隔（与服务启动时 clamp 一致）">间隔</th>
               <th>status</th>
-              <th>last_run_at</th>
+              <th title="每次周期开始时刷新；若单次同步耗时很长，间隔列仍为周期 sleep，实际两轮完成间隔 ≈ 同步耗时 + 间隔">
+                last_run_at
+              </th>
               <th>last_error</th>
               <th>操作</th>
             </tr>