* Change service discovery to relabel jobs to align with
rule expressions in upstream examples
* Use a separate service account for prometheus instead
of granting roles to the namespace's default
* Use a separate service account for node-exporter
* Update node-exporter and kube-state-metrics exporters
