[FIX] distribution.py: Fix computation on multiclass data

[pavlin-policar]

#2680, but without 7d675ba. Apparently force-pushing on a closed PR makes it impossible to re-open it, hence the new PR... @kernc could you take a look?

Issue

The function class_distribution was apparently never used throughout the entire codebase and was never tested, and the parameter ordering was incorrect.

Description of changes

af5a1c7 fixes the mentioned issue. The other commits add various tests, docstrings and code formatting fixes. I've also added a better sample method, similar to the python random.sample. I've also added min and max for ordered categorical variables.

Includes

• Code changes
• Tests
• Documentation

[codecov-io]

Codecov Report

Merging #2903 into master will decrease coverage by <.01%.
The diff coverage is 75.36%.

@@            Coverage Diff             @@
##           master    #2903      +/-   ##
==========================================
- Coverage   82.15%   82.14%   -0.01%     
==========================================
  Files         329      329              
  Lines       56290    56290              
==========================================
- Hits        46244    46240       -4     
- Misses      10046    10050       +4