Implement basic backend cache

[Subject38]

Also paginate jubeat's get_mdata request. For now, this change is very experimental and could probably use some sort of testing. This is not ready to merge but review would be appreciated. The cached decorator is based heavily off the one from here: https://stackoverflow.com/a/16723844 If merged this will close #41 since paginating the database lookup will no longer be required if we can just do one big lookup first. I'm hesitant to add the new lib (pylibmc) to requirements.txt since this should likely be an optional thing to run since it also requires having memcached installed and running.