Design a semi-sparce M matrix for storing the results of moving windows

I think it would be great to allow for the computation of the landscape matrix, M, from moving windows. In the case of none-identical windows this would result in a sparce M matrix. My gut feeling is that this is needed to allow for the computation of connected habitat for sources over targets, see issue #37 