use un-clustered UpgradeStateListener with cache

use un-clustered UpgradeStateListener with cache

doing this with a clustered DTCL only leads to
OptimisticLockFailedException (OLFE) on nodes, but there's really no
need for this to be clustered; it's sufficient for 1 node (leader) to
ract to changes and update the flag.

We also replace the isUpgradeInProgress with a DataObjectCache, because
that (only) does need to use a clustered DTCL to work correctly.

JIRA: GENIUS-207
Change-Id: I08601afb77ca9faaa29e2839aefcdfbcbbb66f5e
Signed-off-by: Michael Vorburger <vorburger@redhat.com>