Optional built-in monitoring

Currently every OCP cluster comes with a fully fledged monitoring stack installed.

With OpenShift being applied to an increasingly varied range of use cases (from MicroShift and SingleNode deployments to very large clusters and Hosted Control Planes) this One Size Fits All approach needs a change.

This initiative was originally titled "Optional CMO". However the requirements indicate that we might still need a CMO-like component to manage some dependencies and provide telemetry data.

The goal is to provide cluster admins to configure whether they want fully fledged monitoring or not. 

What exactly "Optional built-in monitoring" means will likely develop over time. When a cluster admin decides against a fully fledged monitoring stack we want to use a few resources as possible. We could simply not deploy any monitoring components to achieve this. However this is currently not possible since monitoring is depended upon by other functionality (auto-scaling, telemetry). See https://docs.google.com/document/d/1KflBZagmyYBIVloWGKyRi-kJR3TuQmH_WIvDJUryq4w/edit# for more details.

As a first milestone "Optional built-in monitoring" will mean that we use a few resources as possible in order to still support existing dependecies.