Create an example on configuring KSM with scaling techniques

[petewall]

On large clusters, kube-state-metrics is unable to keep up and scrape duration goes above 60s (the scrape interval).

KSM has scaling techniques, which are not trivial to implement (more than just setting replicas > 1). We should include an example that shows how to do this and the tradeoffs.