resource_removed is used in many places to monitor removal of
different kinds of OpenStack resource. It is reasonable to expect a
flavor to be removed in a few seconds where as an image based instance
may take minutes to be removed. With that in mind this change
exposes the retry options used by tenacity allowing the caller to
set reasonable expectations for the resource removal.
The restart_on_changed helper was moved to use the config_change
context manager but the code that checks that the config file is
correctly reverted at the end of the test was accidently left within
the context.
