Bare Metal: When removing worker node groups `kubectl get nodes` output includes removed node(s)

[chrisdoherty4]

Summary When following a specific sequence of tasks that removes a worker node group from a bare meatal cluster the associated nodes still appear in kubectl get nodes output with scheduling disabled.

NAME         STATUS                        ROLES           AGE   VERSION
eksa-dev01   Ready                         control-plane   43m   v1.27.4-eks-cedffd4
eksa-dev02   Ready                         <none>          37m   v1.27.4-eks-cedffd4
eksa-dev03   NotReady,SchedulingDisabled   <none>          11m   v1.27.4-eks-cedffd4

The issue occurs only with the reproduction steps below. It does not occur when the cluster is created with N worker node groups and is upgraded to N-1 groups.

Reproduce

1. Create a bare metal cluster with 1 worker node group.
2. Add a new worker node group, reusing the existing machine config, and upgrade the cluster.
3. Remove the newly added worker node group and upgrade the cluster.
4. Observe the node still appearing in kubectl get nodes output.

Environment: Using EKS-A mainline @ https://github.com/aws/eks-anywhere/commit/95457b086b695ef73f13408120ef55c4854e3bf6