README: add resource recommendation

[brancz]

Based on @matthiasr's comment.

@fabxc @andyxning

---

This change is 

[matthiasr]

that's more generous than what I read out of the tests, I basically meant "start with 200MB, start scaling up if you have more than 100 nodes". but no big harm in being generous I guess?

[brancz]

Agreed. What I wanted to avoid is people giving their kube-state-metric 6mb request and limit. 200mb will get you very far from our findings.

[matthiasr]

hmm but in the way you phrased it, I would expect that I need 206MiB for a 3 node cluster, so I'd probably add even more, and then it does get a bit excessive. What do you think about

Resource usage changes with the size of the cluster. As a general rule, you should allocate

* 200MiB memory
* 0.1 cores

For clusters of more than 100 nodes, allocate at least

* 2MiB memory per node
* 0.001 cores per node

These numbers are based on [scalability tests](https://github.com/kubernetes/kube-state-metrics/issues/124#issuecomment-318394185) at 30 pods per node.

This makes it easier for people with "small" clusters, and the math checks out the same. The minimum CPU request is higher just to make things match up nicely, if someone really cares about reclaiming those 0.09 cores they will want to see what exactly they need.

[brancz]

Fair enough, I like your version better 🙂. Adapted.

[matthiasr]

reviewception