classifier_best_split_function: Get a get_best_split function from a classifier
In mlondschien/hdcd: High - Dimensional Change Point Detection

Returns a closure with formal arguments x, start, end and split_candidates that finds the maximizer of the gain_function given data x, start and end on split_candidates using the optimizer optimizer.

1	classifier_best_split_function(classifier, optimizer, control)

`classifier`	A function with formal arguments `x_train`, `y_train` and `control` that returns class probabilities in a matrix of size nrow(x_train) x k, where k is the number of classes in y_train.
`optimizer`	Optimization method to be used to find an (approximate) of the maximum of the gain function. One of `line_search` for BS or similar, `section_search` for OBS or similar or `two_step_search`.
`control`	Control parameter as returned by hdcd_control