fix error handling in elanmanager-impl workers

fix error handling in elanmanager-impl workers

It should not log confusing ERROR messages if it will retry (which it
currently does).  It should however fail the jobs, and the let the
JobCoordinator log the problem (and increments its failure metrics!) if
the caught RuntimeException is anything else than caused by a
TransactionCommitFailedException (currently that will lead to
addToListenableFutureIfTxException doing nothing and the worker to error
log it, but the JobCoordinator to think the job passed, when actually it
really failed).

This is "take 2" with a self counter proposal to the initial attempt in
change Idf603c6ec8679125a9659378af6627c5fb3eb49e, which would be
abandoned if this was deemed to be a more acceptable solution.

JIRA: NETVIRT-1215
Change-Id: I23be5fae88bf908273c47ade1e74348d1e20dada
Signed-off-by: Michael Vorburger <vorburger@redhat.com>