------------------
* Cleaned up inconsistencies, e.g. now using only computations from statsmodels, actually using T-statistics for computing t-test CIs (rather than Z-statistic). This changes CI slightly for small sample sizes.
* Not using Delta correction for relative CIs anymore, i.e. CI computations are done on the abbsolute scale, and then optionally displayed on relative scale. This is a more consistent approach, since Delta corrections were only applied to CIs and not to p-values. This will give more narrow CIs, with the difference being similar in magnitude to the relative change of the point estimates.
* Added ZTestComputer that relies on statsmodels rather than internal libraries to facilitate open-sourcing
* Added open source licence, copyright headers and code of conduct