Clusters created by the magnum-capi-helm driver do not have a heat stack, so
resize currently fails as it depends on the heat stack for a list of worker
nodes. To fix this, we change to querying the nodegroup information and
use that instead.
We keep the old heat stack way for now, as that is the way to get a list
of worker nodes to select candidate(s) for scaling down. This is not
available in capi driver yet, so capi will return an empty list of
worker nodes.
Change-Id: I0d66fc02a7f2c608a4a0b09b98c343017e04ed41
a5ff9ecb4f