Currently, the sampler and splitter classes differ in their behavior. These should be unified as follows: * .split() ... adds a column called 'split' that assigns the row to a class: 1/train, 2/test, 3/eval * .sample() ... extract a subset of the data