Allow computing orbital graphs of a (semi)group, up to isomorphism

[markuspf]

We currently only compute a set of representatives for isomorphism classes of orbital graphs for permutation groups, and surely that is enough?

The same should be done for transformation semigroups.