add service monitor for grafana

Add SessionAffinity ClusterIP to prometheus service

Adds SessionAffinity ClusterIP to the prometheus service to prevent grafana from querying different instances when loading/updating metrics.

This patch  allows kube-state-metrics to list and watch instances of
poddisruptionbudgets from the policy API group.

Increase CPU limits for small containers to not being throttled as often

the operator spec key is named "externalUrl", rather than "externalURL"

Do not error if no $._config.prometheus.rules are set

Explicitly reference values we are interesting in (jsonnet style)
instead of just following conventions (helm style)

Fix AlertmanagerMembersInconsistent rule