Support starting existing cluster from 0 controller nodes

k0sctl version v0.15.5 on NixOS, k0s 1.27.5 on Rocky 9.

We've setup a 3 node controller+worker cluster. Due to a misconfiguration of the storage extension k0sctl took down all 3 nodes at once (another bug I guess). After that it was unable to start the cluster back up, because it attempted to start the nodes one by one, however the etcd obviously never get's a quorum with 1 out of 3 members. To fix the situation we manually started 2 of the 3 nodes simultaneously, so that etcd would become operational.

So to reproduce:

1. Setup a cluster with 3 controllers
2. Stop all 3 controllers
3. Try applying the k0sctl config

Expected:

k0sctl realizes that not enough controllers are running to get a quorum and decides to start multiple nodes at once.

Actual:

k0sctl tries to start first node which eventually fails.