vsphere-cloud-controller-manger only sets providerID on tainted nodes

[Adam Dyess]

According to upstream docs[0] concerning the cloud-manager indicate that when kubelet on a kubernetes node starts, it should have `--cloud-provider=external` set which will taint the node with

* node.cloudprovider.kubernetes.io/uninitialized=true:NoSchedule

To prevent scheduling workloads on this node until the cloud-provider is started. the vsphere-cloud-controller-manager goes an extra step and doesn't set the providerID and other annotations on a node until it is running and discovers a node with this taint in place. Once it finds the node, it will set its annotations, set the providerID, and remove this taint.

It's possible to start charmed-kubernetes without a cloud-provider charm attached, but once the charm is started and related -- the nodes already enlisted into the cluster do not have this taint. Even restarting kubelet with the new argument `--cloud-provider=external` has no effect on adding this taint because the node is already registered.

In order for the vsphere-cloud-controller-manager to action upon those nodes, that taint must be set:

WORKAROUND:
Taint all the nodes once the vsphere-cloud-controller-manager is running in kube-system
> kubectl taint nodes --all node.cloudprovider.kubernetes.io/uninitialized=true:NoSchedule

[0]: https://kubernetes.io/docs/tasks/administer-cluster/running-cloud-controller/#running-cloud-controller-manager