Whether a pod has unbound volumes influences scheduling decisions and
thus the scale up decisions in cluster autoscaler.

These three new test cases cover:
- a pod with an unbound pvc using late binding -> can scale up
- the same with storage capacity feature enabled -> cannot scale up
  without CSIStorageCapacity
- the same with manually configured CSIStorageCapacity -> can scale up

The assumption that all node labels except for the hostname label can be copied
verbatim does not hold for CSI drivers which manage local storage: those
drivers have a topology label where the value also depends on the hostname. It
might be the same as the Kubernetes hostname, but that cannot be assumed.

To solve this, search/replace with regular expressions can be defined to modify
those labels. This then can be used to inform the autoscaler about available
capacity on new nodes:

   --replace-labels ';^topology.hostpath.csi/node=aks-workerpool.*;topology.hostpath.csi/node=aks-workerpool-template;'

   kubectl apply -f - <<EOF
apiVersion: storage.k8s.io/v1beta1
kind: CSIStorageCapacity
metadata:
  name: aks-workerpool-fast-storage
  namespace: kube-system
capacity: 100Gi
maximumVolumeSize: 100Gi
nodeTopology:
  matchLabels:
    # This never matches a real node, only the node templates
    # inside cluster-autoscaler.
    topology.hostpath.csi/node: aks-workerpool-template
storageClassName: csi-hostpath-fast
EOF

When a new node becomes ready, a CSI driver is not going to be running on it
immediately. This can cause the cluster autoscaler to scale up once more
because of pending pods that can run on that new node once the driver is ready.

The actual check is about CSIStorageCapacity. By comparing the published
information about the new node against the information for a template node, we
can determine whether the CSI driver is done with starting up on the node.

The new CSI processor needs information about existing CSIStorageCapacity
objects in the cluster, just like the scheduler predicate. Both can share the
same informer. For that to work, managing the informer factory must be moved up
the call chain so that the setup code for both can use the same factory.