Add annulus kernel to focal statistics

[chase-dwelle]

In addition to the circular kernel, I would like to have the ability to use an annulus as a mask for calculating spatial statistics.

I am dealing with high resolution datasets and radii in the [km] range. Don't think there would be issues with converting from circle->ellipse at larger radii for an annulus, but correct me if I'm wrong.

Happy to contribute this if the authors feel it's useful.

[brendancol]

@chase-dwelle hey thanks for the suggestion

Would love an annulus kernel. We also just generally need more eyes on the focal module and would love more contributors.

[chase-dwelle]

Thanks, opened up a PR to add the kernel. Probably worth considering abstracting out kernel generation into its own class or set of functions, though I am not sure what other kernels are popular with geospatial analysis.