Add support for a cluster pool auto scaler

[bassam]

We should support a new controller (a cluster autoscaler) that can automatically create Kubernetes clusters on as needed basis, and remove any idle clusters.

The autoscaler would watch Workloads and look for ones that are unschedulable or have reached capacity/limits on existing clusters, and react by creating new Kubernetes clusters from defined pools. For example:

 apiVersion: compute.crossplane.io/v1alpha1
 kind: KubernetesClusterPool
 metadata:
   name: default-pool
   namespace: crossplane-system
 spec:
   minCluster: 1
   maxCluster: 10
   clusterSpec:
      clusterVersion: 1.10
      nodePools:
      - name: default
        machineType: n1-standard-1
        autoScale: true
        minNodes: 1
        maxNodes: 10

[erwbgy]

This is a great idea, but the cluster autoscaler terminology could be confusing because the Kubernetes cluster autoscaler "is a tool that automatically adjusts the size of the Kubernetes cluster" so scales the nodes in an existing cluster, whereas this feature is about scaling clusters rather than nodes. Maybe change the title and description to _cluster pool autoscaler_to match the kind?

[jbw976]

good suggestion @erwbgy! i updated the title to reflect that

[negz]

I don't think this is especially relevant with where Crossplane is at today - we can reopen if and when we invest more in native or day two workload support.