add max_features and tokenizer to CountVectorizer

[marco-cloudflare]

add max_features and tokenizer to CountVectorizer (similar to what's available at sklearn). Note that tokenizer and regex as competing parameters, in case sklearn, it disables regex if you pass a tokenizer and gives you a warning, so here we could think of a single parameter that would encompass both.

another caveat is the serialization of the tokenizer function pointer, the workaround I made was not skip it, allow it to be reset after deserialization and keep a guard that will error if you try to use transform after deserialization without resetting a tokenizer

[codecov]

Codecov Report

Attention: Patch coverage is 54.54545% with 10 lines in your changes missing coverage. Please review.

Project coverage is 35.46%. Comparing base (a30e5f1) to head (31a78fb).
Report is 1 commits behind head on master.

Files with missing lines	Patch %	Lines
...gorithms/linfa-preprocessing/src/countgrams/mod.rs	52.94%	8 Missing ⚠️
...ms/linfa-preprocessing/src/tf_idf_vectorization.rs	50.00%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #376      +/-   ##
==========================================
+ Coverage   34.93%   35.46%   +0.52%     
==========================================
  Files          96       96              
  Lines        6334     6384      +50     
==========================================
+ Hits         2213     2264      +51     
+ Misses       4121     4120       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[relf]

Thanks for your contribution. As you mentioned, it would be great to have only one parameter with an appropriate enum type to unify regexp and tokenizer function.
Could you also add some tests and maybe an example of the tokenizer function you introduce?