Depthlist values must be scaled to sum(read_counts).

[TyberiusPrime]

I was observing all my correlations being between 0.118 and 0.119,
and no gene was classified as 'highly variable' (and VariableFeatures(seurat_obj) was NULL).

Turns out, I was supplying the barcode read counts as depthlist,
but they must be divided by the sum of all read counts.
(If I read get.residiuals source correctly).

I think it's worth adding an assertion that sum(depthlist) must be 1.0 +- epsilon, and another half sentence in the documentation.

Thanks.