Refactor the MultiIndexTree class

the concept of a multi index tree exists (at least) for complete sets of multi indices (in addition with their lexicographical ordering). currently the MultiIndexTree class is only being used for storing intermediary results of this tree representation (node positions, sizes etc.) these results are being used by the divided difference schemes ("nD DDS") and hence also the barycentric transformations.

discuss if this class should be kept alltogether or if the functionality should be included somewhere else.