I propose that we move the adapters for PicketLink and MongoDB into separate Maven
modules. IMO the code in services is already a bit crowded, and with the addition of a
MongoDB implementation it will only get worse. If we add a SPI for these it would be
easier in the future to add more. It would also be possible for users to write their own
if they really need to.
This will also make it possible for us to release M1 with either PicketLink or MongoDB
even if the non-default implementation is not fully working at the time.