[Feature] Sparse Matrix Representation

[costrouc]

Sparse matrix computations were neglected in numpy. I would like to start discussion on where sparse matrix representations fit. As far as hierarchy goes I think that it does not necessarily belong in XND but should possibly exist at a level below the numerical libraries such as gumath. It would fit well because there are plenty of sparse numerical computing libraries out there for both the GPU and CPU that could take advantage of this. I get that dense matrices and space matrices are not exactly compatible.

While I don't know the history some of the representations are in scipy. These formats would all be possible to efficiently represent in XND.