Command used:
operator-sdk add controller \
 --kind=PagerDutyIntegration \
 --api-version=pagerduty.openshift.io/v1alpha1

This doesn't yet watch everything, and it doesn't handle migration of
existing configured clusters to the new way with the
PagerDutyIntegration CRD.

Since a reconciliation request for this controller is the name and
namespace of a PagerDutyIntegration instance, and there might be
multiple instances tied to e.g. a ClusterDeployment instance on which
an event occurs; we want to trigger a reconciliation request for each
PagerDutyIntegration that may be affected.

Before this change there were several places where the names of
SyncSet/ConfigMap/Secret resources were decided, and since those are
more dynamic in this patch set, it makes sense to try to tidy it up a
bit.

Since this patch set changes the way this operator works to reconcile
a new custom resource, we need to be able to get existing
ClusterDeployments with their PagerDuty services into this new
format.

Rather than adding some sort of additional field to the spec that
would only ever be useful once for one CR, this looks for the
following annotation on CRDs: "pd.openshift.io/legacy" with a
non-empty value. This allows us to set this annotation on the OSD
PagerDutyIntegration CR, so that the operator knows that this is the
one to use for migration.

Once it has migrated all of the resources, it will delete the
annotation from the PagerDutyIntegration CR, so that it won't attempt
to migrate on every reconciliation.

This takes the old tests from the ClusterDeployment controller that
was here before, and makes them work with the PagerDutyIntegration
controller.

NOTE: Some of the tests have been commented out until some
dependencies get updated, as the fake k8s client can't handle list
options to a client.List call correctly.

This was a requested change from code review:
https://github.com/openshift/pagerduty-operator/pull/90/files#r448159838