Account for demultiplexing without genetic demultiplexing

[allyhawkins]

Closes #740

Here, I am making a small adjustment to include sample cell estimates for multiplexed libraries when no genetic demultiplexing is run. We already had an argument, demux_method, set up in the script to generate the metadata and QC report, but we were not modifying the default of vireo regardless of whether genetic demultiplexing was run or not.

I added another value to the input channel for the QC report indicating if that sample was demultiplexed using genetic demultiplexing. Then, that gets used to determine the value used for demux_method when running sce_qc_report.nf. If no genetic demultiplexing was used, I'm using hashedDrops. I think we could add a parameter to specify using hashedDrops or HTODemux, but that's probably overkill.

I did originally think we should add a key to the meta object, but if we want to make sure every library has that key in their meta object we need to do it before any quantification happens. But then if mapping is skipped the meta object that is in scpca-meta.json won't have that value. I don't think this is worth making an update to all the scpca-meta.json files, so I thought adding it right before we actually need it made the most sense.

[allyhawkins]

@jashapiro I liked this idea of doing everything within the script so I made some changes to do that.

• I added a list of demux_methods used to the metadata of the filtered SCE object. If we like this change, then I can update the docs accordingly.
• Then in sce_qc_report.R, I pull out those demux_methods and use vireo, if present, to estimate cell counts. If it's not present, then I just grab the first in the list, which should be hashedDrops. I honestly don't think it really matters between hashedDrops and HTODemux which one we use?

This should now be ready for another look.