Deal with single-cell samples lacking data

[arteymix]

Some datasets in GEO only provide data for some of the samples we identify to be single-cell.

It would be unwise to exclude those samples outright as we might have other opportunities to find data elsewhere such as by looking into SRA.

Our data model supports "empty" samples and we use that whenever there are no cells for a given gene. We could use that for every design element. An alternative would be to exclude the sample altogether from the SCD, but that might not play well with the rest of the codebase.

• add flags to ignore those samples when downloading and loading single-cell data
• add empty samples to the single cell vectors (or exclude them from the BAD)
• produce an error if all samples are missing