Cluster Image Registry Operator Should Support Tunable Pod Affinity For Node Placement

1. Proposed title of this feature request
ClusterImageRegistryOperator should support affinity options for nodePlacement

2. What is the nature and description of the request?
ClusterImageRegistryOperator supports specifying nodePlacement options (nodeSelectors and tolerations). It does not currently support the more flexible podAffinity and nodeAffinity options.

3. Why does the customer need this? (List the business requirements here)
The more flexible podAffinity and nodeAffinity options would allow for conditions like preferredDuringSchedulingIgnoredDuringExecution, so that infrastructure nodes could be preferred, but if not available, the pods will still schedule onto other nodes. With how critical these pods are, it is preferred that it be allowed to schedule in many scenarios, even if there are preferences for where it should go if there is capacity.

4. List any affected packages or components.
https://github.com/openshift/cluster-monitoring-operator/blob/eec0457c7449685e8291f6e4b575e25bc3ad8a98/assets/prometheus-k8s/prometheus.yaml#L10-L22