Add input_masks to reproject_and_coadd

Currently we can pass weights to reproject_and_coadd but it might be nice to provide a way to provide boolean masks (which could then easily be generated from data quality arrays).

Start by locating the reproject_and_coadd entry point and tracing how its existing weights are accepted and used. Determine where boolean input masks should enter that flow, then verify that masks can be generated from data-quality arrays and are applied correctly during coaddition.

Written by the indexing model from the issue text.