The dump/restore function works by unrolling all relations in a sequence.
This should work in most situations, but it may well happen that some fringe
cases the resolution will get stuck. Tests have been corrected to work with the
revised version of the backup.

This will allow to create ordered dumps to allow for remote
syncing of the db

Now it is possible, by using the "identifier" property, to define more
than one many-to-many relation between two entites.

By default, the identifier is valued "1", so no need to change existing schemas.

Right now it is still impossible to have two entities access the same relation.