This patch plumbs storage drivers so they all get a cache instance
they can use at their discretion.
I decided to make cache a required param to the initializers, since
any deployment of Marconi of any significance would likely opt to
enable caching anyway, and the cache library has a default in-memory
backend that can be used for simple deployments.
Note also that both data- and control-plane drivers receive a cache
instance, even though the sharding controllers will not use it
(caching will be done in the sharding data driver instead.) I thought
it would be better to pass cache in both cases so we can share test
code and avoid complicating utils.load_storage_driver(). Also, the
control driver may eventually support operations other than sharding;
cache may come in handy then.
Change-Id: I647791af0d7a5914c30cb2489033ec650a455370
Signed-off-by: kgriffs <kurt.griffiths@rackspace.com>
ecce97fd03