Revisit the fill map algebra function

[djohnson729]

It looks like the fill map op adds missing tiles filled with the specified fill value, but it does not replace nodata values in existing tiles with the fill value. We need to evaluate whether or not it should do this.

[djohnson729]

The fill operation takes a fill constant or a fill raster as an argument. It should run through all of the tiles within the bounds of the source tile, and if the source tile exists, for each pixel that has a nodata value in the source, it will replace that value with the specified constant (or with the value of the same pixel in the fill raster if the fill raster argument was specified instead of a constant).

The fillbounds operation works the same except that it also accepts arguments that define the bounds to process (either the coordinates or a raster whose bounds should be used). In this case, rather than running through all the tiles in the bounds of the source image, it runs through all the tiles within the specified bounds and performs the same processing described above.