Fix NetworkAttachments on Cinder Volume

When we deploy multiple cinder volume backends with network attachments
the reconciliation loop is constantly getting called.

This issue happens because each of the cindervolume objects ends up with
the IP addresses of the network attachments of all the different
backends as the labels that these services are the same.

This patch adds a new label with the backend name so that the network
attachments are not merged.

Before the patch we would see both IP Addresses:

NAME                         NETWORKATTACHMENTS                                                                                  STATUS   MESSAGE
cinder-volume-lvm-iscsi      {"openshift-sdn":["10.217.1.74","10.217.1.72"],"openstack/storage":["172.18.0.32","172.18.0.30"]}   True     Setup complete
cinder-volume-lvm-nvme-tcp   {"openshift-sdn":["10.217.1.74","10.217.1.76"],"openstack/storage":["172.18.0.32","172.18.0.30"]}   True     Setup complete

After the patch we would see:

NAME                         NETWORKATTACHMENTS                                                       STATUS   MESSAGE
cinder-volume-lvm-iscsi      {"openshift-sdn":["10.217.0.211"],"openstack/storage":["172.18.0.32"]}   True     Setup complete
cinder-volume-lvm-nvme-tcp   {"openshift-sdn":["10.217.0.210"],"openstack/storage":["172.18.0.30"]}   True     Setup complete

controllers/cindervolume_controller.go

@@ -310,6 +310,7 @@ func (r *CinderVolumeReconciler) reconcileNormal(ctx context.Context, instance *
 	serviceLabels := map[string]string{
 		common.AppSelector:       cinder.ServiceName,
 		common.ComponentSelector: cindervolume.Component,
+		cindervolume.Backend:     instance.Name[len(cindervolume.Component)+1:],
 	}
 
 	//

pkg/cindervolume/const.go

@@ -18,4 +18,5 @@ package cindervolume
 const (
 	// Component -
 	Component = "cinder-volume"
+	Backend   = "backend"
 )