Consistent handling of sparse data in fit and transform

[lmcinnes]

As per #81 there are some issues with how transform manages to deal with sparse/dense data. This should be made consistent so that you can (ideally) use any mix of sparse and dense as required cleanly with both fit and transform.

[arita37]

Hi, Still showing error for umap-learn==0.5.1 should we use dense for .transform ? .fit works well with sparse