Follow-up from "Finalise apb": deprovisioning should not fail if resources have already been deleted

The following discussion from !21 should be addressed:

• @alossent started a discussion:

  Deprovisioning should NOT fail if the resources have already been deleted. This actually applies to all deprovisioning steps:

  • if swarm cluster is gone already, do not fail but just skip stopping gitlab-runner on each node and removing the cluster
  • do not fail when stopping gitlab-runner on each node one by one if a node is not responding: it's OK, we don't need to stop gitlab-runner on that node since the node is down
  • also do not fail when removing the headless service if it's gone already (this should be the case already when using k8s: {state: absent})

  This is going to require some changes and it's not critical for this MR, but it does create a risk of leaving resources behind. So moving it to a separate issue, to be addressed in a separate MR.