implement Multimeasure class

[gregmedlock]

We've started implementation of a Multimeasure class, which holds multiple AnnData objects (https://github.com/theislab/anndata). These AnnData objects are the primary object used by scanpy (https://github.com/theislab/scanpy) for analysis and visualization of single-cell RNA seq data.

[jfx319]

turns out for the ingestion of data, we may not need to call scipy.sparse ourselves (may actually want to use or extend the existing scanpy/anndata for consistency):

the anndata object specs is worth a skim through, to keep this in mind. read_mtx (PS: @nsheff they implement scipy.sparse and choose the csr variant after all)

anndata implements a transpose() that also interchanges the observations/variables

• could use this concept to first load both the mtx & annotations in the wrong way, and then call the transpose to flip it
• or load the mtx without annotations, then transpose, then manually set the annotations explicitly (more logical to a reader, despite more steps)