Lack of support for sparse matrices makes scNym unusable for large datasets

[Wainberg]

scNym requires log(CPMs) rather than raw counts. This requirement makes scNym unusable for large datasets, since log(CPMs) require a dense matrix, and dense matrices usually can't fit in memory for datasets with more than a few hundred thousand cells.