Currently, cinder client does not retry when connections to cinder
service is refused. There are many legitimate scenarios under which
retry should be attempted:
1) cinder service being restarted;
2) cinder service is running on multiple API nodes behind a LB, which
might be temporarily overwhelmed or being maintained.
In any scenario, retry with a backoff timer does not seem to hurt.
Change-Id: I3c290c59fa67262c4a3473815b4380ee39e24332
Closes-Bug: 1347843
fd8aa39458