[FIX] preprocess: Use default tokenizer when None

[lanzagar]

Issue

A preprocessor can be constructed with tokenizer=None (actually, this is the default!).
In this case preprocessing produces a single token with the complete text. This can cause many problems and is probably never desired/useful.

Description of changes

1. If tokenizer=None, use the same default tokenizer as is used in base_preprocessor, which is called when accessing Corpus.tokens without previous explicit preprocessing.
2. Move base_preprocessor into the preprocess.py module, right after the class Preprocess (of which it is an instance).

Includes

• Code changes
• Tests
• Documentation

[codecov-io]

Codecov Report

Merging #294 into master will increase coverage by 0.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master     #294      +/-   ##
==========================================
+ Coverage   85.02%   85.04%   +0.01%     
==========================================
  Files          34       34              
  Lines        1823     1825       +2     
  Branches      333      333              
==========================================
+ Hits         1550     1552       +2     
  Misses        238      238              
  Partials       35       35

[nikicc]

Should we put this into the documentation also? Something like tokenizer (BaseTokenizer or None): tokenizes string, uses WordPunctTokenizer when None is given?