Run immediate health probe after jail deactivation

After deactivation the endpoint now calls _run_probe to flush the
cached server status immediately, matching the activate_jail behaviour
added in Task 5. Without this, the dashboard active-jail count could
remain stale for up to 30 s after a deactivation reload.

- config.py: capture result, await _run_probe, return result
- test_config.py: add test_deactivate_triggers_health_probe; fix 3
  pre-existing UP017 ruff warnings (datetime.UTC alias)
- test_health.py: update test to assert the new fail2ban field

backend/tests/test_routers/test_health.py

@@ -13,10 +13,11 @@ async def test_health_check_returns_200(client: AsyncClient) -> None:
 
 @pytest.mark.asyncio
 async def test_health_check_returns_ok_status(client: AsyncClient) -> None:
-    """``GET /api/health`` must return ``{"status": "ok"}``."""
+    """``GET /api/health`` must contain ``status: ok`` and a ``fail2ban`` field."""
     response = await client.get("/api/health")
     data: dict[str, str] = response.json()
-    assert data == {"status": "ok"}
+    assert data["status"] == "ok"
+    assert data["fail2ban"] in ("online", "offline")
 
 
 @pytest.mark.asyncio