Fix call to preprocess_split() in generated documentation.

[rcurtin]

This is a fix for #3556, in the Python bindings. The generated documentation would print a line like

X_test, y_test, X_train, y_train = preprocess_split(input_=X, input_labels=y)

but preprocess_split() returns a Python dict with several members. Instead, this is what's necessary to get the training/test datasets:

d = preprocessd_split(input_=X, input_labels=y)
X_train = d['training']
y_train = d['training_labels']
X_test = d['test']
y_test = d['test_labels']

These changes update the function that prints this documentation to use the second version. I tested that the generated code runs successfully.

[NippunSharma]

LGTM 👍

[NippunSharma]

As a side note: should we also look into changing what preprocess_split returns and change that to individual matrices. That seems to be a bit more "pythonic" way of doing it?

[mlpack-bot]

Second approval provided automatically after 24 hours. 👍

[rcurtin]

@NippunSharma agreed, it would be great if we could make the preprocess_split() API cleaner, but I am not sure of a nice automated way to do that with our binding generation system. If you have ideas I am all ears 😄