Wait for all nodes being added before running ansible

[jprovaznik]

Openshift compute ndoes are members of AutoscalingGroup in heat
stack, when each member is created it registers itself on
bastion node ("deployment_bastion_node_add" SW deployment in
templates), then a separate step ("deployment_run_ansible"
in node.yaml) configures openshift on all nodes which are
already registered.

It may happen that deployment_run_ansible is trigerred by
some node before all nodes are registered (by deployment_bastion_node_add)
which causes that then nodes are added in multiple ansible
runs/batches. It doesn't break setup and basically is
nothing bad with it but it makes creation or scale up
slower. It would be better to wait until deployment_bastion_node_add
is done for all nodes and then run deployment_run_ansible
only once for all nodes (AutoscalingGroup doesn't allow to do this easily though).

This patch uses node_count param to get desired number of nodes,
this can not work with autoscaling though so it's used only for
not-autoscaling create/update scenario.

[jprovaznik]

Please do not merge into a branch for GA release (this is not a blocker, I would prefer to have this merged in a next release)

[markllama]

@jprovaznik We'd like to test and merge this now but it requires a rebase. If you can do that quickly we can get it merged.

[jprovaznik]

@markllama I've rebased the patch, but please pay extra attention to assure it still works - I did just rebase of the patch, no teesting.

[tomassedovic]

Thanks!