add weights in pseudbulkDGE for edgeR analysis

[koenvandenberge]

Hi,

In some cases, one might want to add weights to the dispersion estimation and fitting of the edgeR model when using the pseudobulkDGE function. In my case this would be sample-level weights (much like voom with sample quality weights), but I imagine that observation-level weights may also be of interest, so I have allowed a matrix of weights that is of same dimension as the expression matrix. This seems to be possible using only a minor adjustment.

Let me know what you think.