Proposal for user cluster nodes remediation

[mfranczy]

As a cluster admin I would like to have a mechanism that will allow me to safely drain bare metal nodes with respect to user cluster nodes running on top (virtual machines).

Things we should check/change:

• Eviction strategy of VMs should be set to External
• We should respect run strategy manual (in that case do nothing and leave it to users)
• Check possibility of live migration - probably we will have to follow re-create VM nodes strategy
• Allow to trigger a user cluster node remediation by setting annotation (mostly for tests)

Acceptance criteria:

• A document that describe how to:
  • Support user cluster workload eviction caused by bare-metal node draining
  • How to integrate the mechanism with machine controller's node eviction
  • Design of e2e tests that covers the feature
• Issues in the epic will be created based on the document