Very small bias in RR estimate for auto-correlation functions

The following is a very minor point. There's a nice notebook by Lehman Garrison showing that the naive analytic RR estimator, RR = N dV N / V, has a slight bias. Instead, it should be RR = (N - 1) dV N / V. Halotools seems to be using the naive RR estimator with the slight bias.

https://nbviewer.jupyter.org/gist/lgarrison/1efabe4430429996733a9d29397423d2

Should be trivial to correct in halotools. It's a very minor point because N is generally very, very large.