Remove numpy from default PythonInterpreter and potentially introduce PythonNumpyInterpreter

Right now only Python uses third party library (specifically numpy) for linear algebra. This is inconsistent with:

• our mission “with zero dependencies”
• other languages.

The first step was to drop numpy from cases without vectors, implemented in PR https://github.com/BayesWitnesses/m2cgen/pull/111 .

As the second step I want to suggest dropping numpy altogether from PythonInterpreter and potentially implement PythonNumpyInterpreter to use in cases where it would be beneficial.

As for the user API I see 2 options:

1. Adding a new method export_to_python_with_numpy
2. Adding a arameter with_numpy to an existing export_to_python method which would be False by default.

I personally think first option is better as users would have higher chances of noticing extra method than extra parameter with a default value.

[krinart]

My intuition was that one third-party library for linear algebra will be enough in any language

This is how assumptions work. They always seem to make sense until they don’t 😃

In general I always try to make as few assumptions as possible, as every assumption creates a dependency. In this case we would create an artificial dependency “Every language will have only one implementation with linear algebra library”. It seems like it makes sense (and it probably does), but why would we bring this unnecessary limitation?

Especially given that we are talking about our API, which should arguably be the most though-out component as that is exactly what users depend on and making backward incompatible changes in a well established API has never been an easy problem.

[izeigerman]

@StrikerRUS Yeah, I agree. I see no benefits either. I think we should close this. Thank you!