Use AnnData instead of CSV Files

[ilan-gold]

As far as I can tell, most of the outputs are per-observation files/information (like expression per nuclei, boundary, membrane etc. per cell) which fits well into the AnnData model i.e each obs variable would be a different attribute of the cell, like antigen expression (i.e the var variable) within the membrane or polygonal spatial coordinates per segmentation per cell nucleus. I think this could be a good way to standardize outputs and make it easy for future users to use the outputs in a standardized manner via the AnnData API. @mruffalo thoughts? Just spitballing here....we convert the files to AnnData to some degree anyway for visualization because it's so easy to consume!

[ilan-gold]

See: https://anndata.readthedocs.io/