Integrate scanpy for more efficient clustering and general functionality

[adamgayoso]

Scanpy backend implemented

• use Scanpy Louvain clustering
• adds tqdm for progress bar
• validated on internal datasets
• default params run method as is in version 2.4.0
• updates umap plot by showing predicted singlets in grey, doublets in black

[adamgayoso]

We could add back phonograph support as follows

import scanpy.external as sce
result = sce.tl.phenograph(adata.obsm['X_pca'], k = 30)

In fact we should make it an option so that we can always use the original method. This would also involve making the standard scaling an option.

[adamgayoso]

We could add back phonograph support as follows

import scanpy.external as sce
result = sce.tl.phenograph(adata.obsm['X_pca'], k = 30)

In fact we should make it an option so that we can always use the original method. This would also involve making the standard scaling an option.

I added this, but not through the external scanpy module. I added back what we had before.

[adamgayoso]

@JonathanShor This is good to squash into dev. The only thing that is missing is that your PhenoGraph actually doesn't have a verbose or verbosity argument. So, if you choose to use_phenograph and verbose is False then Pheno will still print stuff out.