Aplr

Fixed a minor bug that sometimes added unnecessary terms to the model.

Fixed a minor (and for practical purposes insignificant) inconsistency w.r.t. rounding errors.

Fixed a bug that caused less interaction terms than intended to be considered for model inclusion.

Reverted the default learning rate for APLRClassifier to 0.1.
Updated the documentation.

Introduced bugs in this version and yanked it as a consequence.

**Changes:**
- Added the constructor parameter _early_stopping_rounds_ with a default value of 500, meaning that if the validation loss does not improve during 500 boosting steps then boosting is aborted to save time. Due to _early_stopping_rounds_ it may make sense to try higher values of _m_ (max number of boosting steps).
- Updated documentation.
- Changed default values of _m_ for _APLRRegressor_ and _APLRClassifier_ and _v_ (learning rate) for _APLRClassifier_.