[WIP] Start rewriting the in-mem storage as a backend for cassandra.
Something I wanted to do for a while, I think it will lead to simpler code in the in-mem storage; and less duplication/differences between cass and in-mem.
But I'm still not sure it's a good idea so I'm not spending too much time on it
Migrated from D3194 (view on Phabricator)