The parts of the agent wait-for bootstrap-complete and agent wait-for cluster-complete commands that wait for the bootstrap configmap and the clusterversion (respectively) should be functionally identical to the existing code for the IPI wait-for bootstrap-complete and wait-for cluster-complete commands. (The code that wraps this is, however, necessarily different.)

The current code to do this in the agent commands is a duplicate implementation, and it polls the Kube API instead of using a watch. We should refactor the IPI implementation to the extent necessary that we can just call it instead of reimplementing it.