MachineSet implementation for scaling up & down

[rsdcastro]

Some thoughts/questions:

• Choose a specific instance to remove rather than arbitrary instance
• Does scaling down use any smart logic to pick the node to remove?
• Does scaling down drain nodes by default? (But allow removing machines w/o draining)

[roberthbailey]

We should definitely support removing a specific machine from the set. This isn't supported today in replica sets so there isn't precedence in that API (although customers have asked for it) but it is required by the cluster autoscaler.

Ideally scaling down would choose the best node to remove from the cluster. The cluster autoscaler has code that does this, so it might be possible to reuse that here. For the initial implementation (alpha) we can just pick one randomly.

I'm of the opinion that scaling down should drain by default but allow a "force delete" which skips draining.

[rsdcastro]

This issue was moved to kubernetes-sigs/cluster-api#27