selecting the number and location of background points

[goldingn]

a process module to do this, based on variability in the chosen covariates, would be awesome!

@rvalavi has some ideas for how to select the optimal number of points based on this. K-means clustering might be a good way of efficiently selecting locations of background points (assuming reporting biases are taken care of in some other way)