Support an Un-managed Slurm Controller

[vpenso]

>>> cat sind.yaml 
kind: Cluster
name: default
nodes:
  - role: controller
    managed: false

>>> sind create cluster --config sind.yaml 
11:31:55.738 ERRO managed is only valid for worker nodes, not "controller"

Support managed: false for the controller, which makes all other nodes un-managed implicitly. This should be very useful to debug issues by running slurmctld -Dvvvvvv manually, while working on the configuration in /etc/slurm within the container.

[dennisklein]

I had a related thought when evaluating the create/delete worker CLI: Should we also generalize it or add a create/delete node --role ... as well which would also make those nodes unmanaged?

[vpenso]

I would like to see a sind create controller <cluster> --unmanaged to enable primary & backup controller configurations as well.