ServiceLB DaemonSet keeps stale nodeSelector after the enablelb label is removed

Environmental Info: K3s Version: v1.36.4+k3s1

Cluster Configuration: 1 server

Describe the bug: When you use the svccontroller.k3s.cattle.io/enablelb label to restrict ServiceLB to specific nodes, and then later remove that label from the last labeled node, the svclb DaemonSet keeps its nodeSelector. It never goes back to running on all nodes the way it does by default, so the LoadBalancer can end up with no pods.

Steps To Reproduce:

• Installed K3s: curl -sfL https://get.k3s.io | sh -
• Create a LoadBalancer service (or use an existing one).
• Confirm the svclb DaemonSet has no nodeSelector: kubectl -n kube-system get ds -o custom-columns=NAME:.metadata.name,SEL:.spec.template.spec.nodeSelector
• Label the node: kubectl label node NODENAME svccontroller.k3s.cattle.io/enablelb=true → the DaemonSet gets nodeSelector svccontroller.k3s.cattle.io/enablelb=true
• Remove the label: kubectl label node NODENAME svccontroller.k3s.cattle.io/enablelb-

Expected behavior: With no node carrying the enablelb label, ServiceLB returns to its default of running on all nodes — the DaemonSet nodeSelector is cleared.

Actual behavior: The DaemonSet keeps nodeSelector svccontroller.k3s.cattle.io/enablelb=true. It stays that way indefinitely, and a full k3s restart does not clear it.

Additional context / logs: Looks like onChangeNode in pkg/cloudprovider/servicelb.go returns early when the changed node doesn't have the enablelb label, so updateDaemonSets() never runs on the removal. Possibly the same underlying issue as k3s-io/klipper-lb#73.