Multiple Comparisons

In PR #849 I merged Tyler's enhancements to tukeyhsd into master (see also #809 #789 #824 )

some things left to do

• another round of proofreading docstrings, and cleaner example script
• add p-values to tukeyhsd results
• refactoring, some comments in the PRs and issue
• move out of sandbox
• unit tests for multicomp module (currently only tukeyhsd is really verified)

further enhancements

• plot for all pairs
• other pairwise comparisons (Dunnet should be easy, proportions_pairwise needs better results class)
• more is a never-ending story https://github.com/statsmodels/statsmodels/wiki/SMEP:-Multi-Comparisons