IPTW for Mediation Analyses

[pzivich]

I am not particularly enthusiastic about adding mediation analysis, but I know people like to conduct those analyses. I am only planning on implementing natural direct effects (no support for controlled direct effects, since I don't like the cross-worlds assumption). I have written SAS code for another project to run a mediation analysis with IPTW. It is fairly easy to integrate. Rather than bloating the IPTW class, I plan on making a separate class (maybe MediationIPTW).

This might mean I should revisit #39 and adding mediation for TMLE (also add a MediationGFormula).