Flipping the sense of fillnodata's mask arg

[sgillies]

Looking more deeply into fillnodata(), I'd like to flip the sense of the mask to match what we get when we read a masked array: True for nodata pixels. Doing so makes usage easier, like this:

arr = dataset.read(1, masked=True)
filled = fillnodata(arr.data, arr.mask)

whereas right now you have to pass ~arr.mask.

Even easier would be if fillnodata() checks the input image for a mask:

arr = dataset.read(1, masked=True)
filled = fillnodata(arr) # arr.mask used by default if available