fix(watcher): add delay on success path to prevent busy loop

When RetryWatcher's result channel closes without an error (e.g., after
exhausting internal retries), the outer loop retried immediately with no
delay, causing high CPU usage. Restore 1-second cooldown on success path
to match previous behavior and prevent reconnection storms.

Co-Authored-By: Composer <noreply@cursor.com>

Author: cursoragent

internal/agent/watcher_events.go

@@ -252,6 +252,7 @@ func watchEventsInNamespaceForever(ctx context.Context, config *rest.Config, nam
 				time.Sleep(b.duration())
 			} else {
 				b.reset()
+				time.Sleep(time.Second)
 			}
 			watchSince = time.Now()
 		}

internal/agent/watcher_pods.go

@@ -207,6 +207,7 @@ func watchPodsInNamespaceForever(ctx context.Context, config *rest.Config, names
 				time.Sleep(b.duration())
 			} else {
 				b.reset()
+				time.Sleep(time.Second)
 			}
 		}
 	}