Establish guidance on cluster Node sizing

[craigtracey]

We should provide guidance for how customers decide on the appropriate Node sizings.

• What do we recommend for instance sizing? Why would I choose one size over another?
• Node sizing affects your blast radius when hardware fails/performs lifecycle operations
• How many Pods will fit on a Node?
• A consideration: each Node requires some control plane Pods to be running
• Are there other considerations when choosing Node sizes?
• Description of NodeAllocatable and its impact on sizing.
• How can customers ensure that they are efficiently operating these Nodes (ie. mostly full to reduce cost, but also enough slack to handle new workload requirements)

[srivar]

Adding Ian to this as well, as similar query has come in on AWS side.

[z4ce]

Node sizing is basically the same question as "how big of a server do I need to run linux?" I think your discussion of blast radius and resource allocation is likely the right way to go here. There should also be some bare minimum to run TKG with recommended extensions in a reference architecture scenario (high availability) -- everything above that depends on use case.