Gather prior art on performance distribution comparison

This issue can be closed with a set of existing methods applied in answering the following questions:

- Are two samples coming from the same disribution? 
- How to detect stochastic dominance? 
- Did the modality change? 
- How many modes? 
- Which quantile estimator to use? 