fix: tolerate offline nodes when controlling forward services and editing tunnels

- controlForwardServices: skip offline nodes instead of failing entire operation,
  so forward pause/resume/delete works when some entry nodes are offline
- onNodeOnline: always sync forward state on node reconnect (not just post-upgrade),
  so forwards that changed status while a node was offline get synced
- add ListForwardIDsByNode repo method to sync all forwards (including paused)
- tunnel edit UI: allow deselecting already-selected offline nodes in
  entry/chain/exit selectors, matching the backend's existing tolerance

go-backend/internal/http/handler/control_plane.go

@@ -474,6 +474,9 @@ func (h *Handler) controlForwardServices(forward *forwardRecord, commandType str
 
 		nodeHandled, lastNotFoundErr, err := h.controlForwardServicesOnNode(fp.NodeID, bases, commandType)
 		if err != nil {
+			if isNodeOfflineOrTimeoutError(err) {
+				continue
+			}
 			return err
 		}