Example: B-test null distributions and approximations

[sejdino]

This is an entrance task (and part of the project) of http://www.shogun-toolbox.org/page/Events/gsoc2014_ideas#variable_interactions

See B-test: A Non-parametric, Low Variance Kernel Two-sample Test

Write a Python example reproducing the plots in Figure 1 of the B-test paper above. Example should contain the comparison of the histogram of the null distribution obtained using the permutation method with its gaussian approximation for varying values of B. Diagnostics from Python (such as Q-Q plots) can be used to show that gaussian approximation is valid only for small values of B.

This task depends on #1928
Contact @karlnapf or @sejdino with questions