Sparse abundance counts refactor

[Elliot-D-Hill]

Implementing Abundance counts as a sparse matrix would save memory, computation, remove the need to reindex the similarity matrix, and simplify species_ordering handling.

Sketch of the implementation:

1. Take species_ordering from similarity matrix.
2. Pivot counts, but before constructing the abundance matrix, sort the species indices in the same order as species_ordering.
3. Construct sparse abundance matrix from sorted pivot rows.
4. Clean up old code: remove unused species_subset arguments (but keep in make_metacommunity)

[Elliot-D-Hill]

After some experimentation, it looks like we will likely not see significant speed improvement for most cases, so I am closing this issue for now.