Change computation of role hierarchy in textbook calculus

[stefborg]

To mirror the design of CR3 in the textbook calculus, maybe we should adapt the computation of the role hierarchy as follows?

nf:subProp(?R,?T) :- nf:subProp(?R,?S), nf:subProp(?S,?T) .

(i.e., don't require that the first role inclusion is a directSubProp)