Ignored concepts/properties not taken into account by refinement operator

[LorenzBuehmann]

Problem:

Class expression learning algorithms derived from AbstractCELA allow to set classes and object/data properties that should be ignored during learning. Those ignored entities are only considered during the initialization of the hierarchies in initClassHierarchy() initObjectPropertyHierarchy() initDataPropertyHierarchy()

Those hierarchy objects are put into the refinement operator. The problem is that for some tasks only the underlying reasoner is used:

if(useNumericDatatypes) {
            Set<OWLDataProperty> numericDPs = reasoner.getNumericDataProperties();

or

if(useDataHasValueConstructor) {
            Set<OWLDataProperty> stringDPs = reasoner.getStringDatatypeProperties();

and even in the init() method.

This can lead to solutions that can contain ignored entities, thus, should not be returned and also cause a larger search space than necessary.

[SimonBin]

hopefully fixed by #48