Reconcile differences between methods to generate cell counts

[theathorn]

Lon's old method vs current method.

[hannes-ucsc]

Use .attic/cell_counts.json and make report to determine if there actually are discrepancies.

[GPelayo]

This is a table of all the discrepancies between .attic/cell_counts.json and make report.

Accession ID	.attic/cell_counts.json	make report
GSE44183	48	47
GSE75659	1318	1316
GSE99795	147	0

@hannes-ucsc Also, GSE95435 has counts in .attic/cell_counts.json but has no spreadsheet associated with it in this repo. GSE95435 only appears in the geo_series_accessions field of spreadsheets/existing/GSE114374_project.json

[hannes-ucsc]

The off-by-ones are fixed in #142. I looked at the geo files in GSE99795 and hesitate to agree with @DailyDreaming's call to count transcripts as cells. We'll stick with our call not to consider that project's matrix. That leaves GSE95435 to resolve.