Document AnnData.X type requirements

[JakeSAI]

While running our datasets through the tc pipeline to test clustering, a small issue was found where the normalize() function fails on /allen/aibs/technology/transcriptomics/datasets/small/smalldata_dense.h5ad, and project() fails on compressed representations like /allen/aibs/technology/transcriptomics/datasets/small/smalldata_sparse_compressed.h5ad. Handling all types that anndata accepts for X seems unnecessary, but we need to establish, document, and tweak implementation around specific dense and sparse data types.

Tasks:

• Decide on type for dense anndata X - ndarray
• Decide on type for sparse anndata X - scipy.sparse.csr
• Document type requirements
• Fix normalize dense handling
• Fix project sparse handling

Validation:

• Run dense copy of small dataset through full pipeline
• Run sparse copy of small dataset through full pipeline