Provide a more fine grained way to configure resource requests/limits in OLM deployments

[duritong]

The way through https://github.com/operator-framework/operator-lifecycle-manager/blob/master/doc/design/subscription-config.md#resources has the drawback that it configures this for each container.

Meaning if you configure 1 CPU as a request, it actually requires 2, since kube-rbac-proxy will also get a request for 1 cpu, which is quite tooo much.

Meaning with deployments through OLM you cannot currently scale the manager pod only.