Function to rarefy point data in any number of dimensions. The goal here is to take a large data set and reduce it in size in such a way as to approximately maximize the difference between points. For instance, if you have 2000 points but suspect a lot of spatial autocorrelation between them, you can pass in your data frame, the names (or indices) of the lat/lon columns, and the number 200, and you get back 200 points from your original data set that are chosen to be as different from each other as possible given a randomly chosen starting point