We can just simplify the usage of an sqlalchemy if we just use
(we already have our own ORM like objects anyway) sqlalchemy core
in the first place and have a very tiny layer that converts back and
forth from our very limited object model that we use in our
persistence layer.
This change makes that adjustment, which makes it easier to read
and understand the actions the sqlalchemy backend is doing when
saving, reading and updating data, and avoids yet another layer
that isn't useful for our purposes anyway.
Change-Id: I911c509f65e7845aee86fed1622eaa56970741f2
687ec91379