Allow fractional pixels when generating input shapefile masks

[jonschwenk]

The ability to create masks from input shapefiles is very welcome, but we ran into an issue that I think can be fairly easily resolved. It seems that the masking procedure considers each pixel either "all in" or "all out", i.e. binary. In the case where a polygon is on the order of the pixel size of the raster being sampled, e.g.

,

it would be great for the mask pixels to take fractional values (i.e. not binary), where the value represents the fraction of the pixel that is covered by the polygon. Then (if not already implemented), the mask can be used as a weights grid when doing the spatial aggregations for any raster.

We have a number of smaller watersheds that might overlap 2-4 pixels for coarser datasets and would appreciate this kind of precise zonal statistics.