test: fix flaky test by longer k8s node checks and retries

[utkuozdemir]

Fix the flakiness in ReplaceControlPlanes test that was introduced by the recently added KubernetesNodeAuditController. This controller changed the Kubernetes node deletion logic to not block the ClusterMachineTeardownController for node deletions, as itself would take care of it later.

This causes the replaced control plane node in this test to be removed with a delay, as it is now removed by the audit controller after a grace period.

Improve the assertion between k8s<>Omni nodes by:

• not only asserting the number of nodes but also their names
• adding retries to this assertion, so it would keep trying until the audit controller does its job and removes the node from Kubernetes
• increasing its timeout to give the audit controller enough time (as it gives nodes a grace period of a minute) to remove the node

[utkuozdemir]

/m