fix(update): stop blocking event loop during rebuild + fix infinite spinner

The update endpoint ran the entire git pull + docker build (up to 10 min)
synchronously inside the request handler, blocking the whole server for
everyone while it ran. Separately, the frontend spinner was only hidden on
error, never on success, so it spun forever even when the update worked.

- Run the update in a background thread; the request returns immediately
- Add GET /settings/update/status for progress polling (backup/pull/build/restart)
- Frontend polls status, then waits for the app to come back after the
  container restart, and shows a clear done/timeout message instead of an
  endless spinner

Co-Authored-By: Claude Sonnet 5 <[email protected]>

app/routers/settings.py

@@ -4,6 +4,7 @@ There is no .env file. Every setting lives in the ``system_config`` table
 (singleton row with id=1) and is editable via the Web UI settings page.
 """
 
+import asyncio
 import logging
 import os
 import shutil
@@ -358,13 +359,15 @@ async def trigger_update(
     current_user: User = Depends(get_current_user),
     db: Session = Depends(get_db),
 ):
-    """Backup the database, git pull the latest code, and rebuild the container.
+    """Kick off backup + git pull + container rebuild in the background.
+
+    Returns immediately — the actual work (which can take several minutes,
+    especially the ``--no-cache`` image build) runs in a background thread so
+    it doesn't block this request or any other user's requests while it
+    runs. Progress can be polled via GET /settings/update/status until the
+    container restarts with the new version.
 
-    The rebuild is fire-and-forget — the app will restart in ~60 seconds.
     Only admin users may trigger an update.
-
-    Returns:
-        Dict with ok, message, and backup path.
     """
     if getattr(current_user, "role", "admin") != "admin":
         raise HTTPException(
@@ -383,11 +386,37 @@ async def trigger_update(
             detail="git_repo_url is not configured in settings.",
         )
 
-    result = update_service.trigger_update(config, DATABASE_PATH)
-    if not result.get("ok"):
+    current_status = update_service.get_update_status()
+    if current_status.get("state") == "running":
         raise HTTPException(
-            status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
-            detail=result.get("message", "Update failed."),
+            status_code=status.HTTP_409_CONFLICT,
+            detail="An update is already in progress.",
         )
+
+    # Snapshot the only fields trigger_update() needs — avoids passing a
+    # SQLAlchemy instance into a background thread after this request's
+    # session may already be closed.
+    class _ConfigSnapshot:
+        git_repo_url = config.git_repo_url
+        git_branch = config.git_branch
+        git_token = config.git_token
+
+    asyncio.create_task(asyncio.to_thread(update_service.trigger_update, _ConfigSnapshot(), DATABASE_PATH))
     logger.info("Update triggered by %s.", current_user.username)
-    return result
+    return {
+        "ok": True,
+        "message": "Update gestartet. Dies kann mehrere Minuten dauern — Fortschritt via Status sichtbar.",
+    }
+
+
+@router.get("/update/status")
+async def update_status(
+    current_user: User = Depends(get_current_user),
+):
+    """Return progress of the currently running (or last) update.
+
+    Note: once the container restarts mid-update, this endpoint stops
+    responding for a few seconds — that itself is a signal the swap is
+    happening. The frontend falls back to polling for the app coming back up.
+    """
+    return update_service.get_update_status()