Expanded cell filtering for datasets with protein data

[llumdi]

When citeseq = True and both adata (gene counts) + adata_prot (antibody counts) are loaded, we should include some cell filtering based on the protein counts metric. I already spotted 2 scenarios requiring filtering:

1. If a cell has 0 counts for all antibodies, normalization function bc.st.clr_normalize(adata_prot, os.path.join(results_folder_citeseq, 'citeseq')) will raise this error:

ValueError: Input matrix cannot have rows with all zeros

Suggestion: add a param min_protein_counts = 1, so that cells with antibody_counts sum = 0 will be excluded

2. During sample prep antibodies can aggregate leading to exceptionally high counts in few cells (see this 10X note). 1 solution is to work with the filtered cellranger matrix (which already excludes such cells), another solution (not exclusive) is to include a parameter max_protein_counts to exclude cells above a threshold.

Add in this issue other possible scenarios to be considered.

Note: the filtering should be performed at the beginning of besca, so that the gene analysis part has the same cells than the protein analysis.