Implement `delete machine --skip-drain-delete`

[dlipovetsky]

When the cluster control plane is in a degraded state, a node cannot be drained or deleted, but the user should have some way to delete the machine.