Allow for multiple kube-state-metrics series

This can happen if you run multiple replicas, or if you redeploy kube-state-metrics.
In either case, prometheus records multiple metrics with the instance ip in, and the dashboard fails.

Use aggregation functions to get sensible output in either case

git-subtree-dir: contrib/kube-prometheus
git-subtree-mainline: 050ca21276696c8603375c699513ec487301ed62
git-subtree-split: 81c0d2f4