ClusterResources should take the operator and controller names separately

[nightkr]

Currently, ClusterResources takes a single manager argument that must be unique per controller. The documentation suggests identifying both the controller and operator names (foo-operator_foo-controller), but in practice many single-controller operators just use the operator name (foo-operator).

This is not only inconsistent across the platform, but can lead to orphaned resources during upgrades that add more controllers (necessitating a switch to the expanded format). Additionally, it's generally a convention that un-namedspaced identifiers are reserved for the cluster administrators and users, third-party software should use the ns/id format for string keys (such as field managers or labels).

Acceptance

• ClusterResources::new takes separate arguments for the operator and controller names
• labels::get_recommended_labels takes separate arguments for the operator and controller names
• managed-by labels and field managers follow the operator/controller format
• The documentation is amended to encourage FQDN-style operator names (for example: zookeeper.stackable.tech rather than zookeeper-operator)

[sbernauer]

Comment on the PR:

I would combine both tickets into one PR because they create a breaking change at the same place. If we would merge this PR and not immediately the PR for #491 then we would force people to adapt their code twice. But I would also not implement #491 hastily without a thourough refinement.