[4] Implement faster Hermite functions in Cython

[MothNik]

This pull request

🦾 Features/Enhancements

• implements a multi-threaded Cython function to compute the Hermite functions way faster than possible with NumPy and Numba so far (see below)
• adds a function to only evaluate single Hermite functions with a complex integral rather than a recursion. This is way faster for high orders, but only works for single functions and not a full basis
• "deprecates" the old NumPy- and Numba-implementations but still keeps them for comparison
• integrates all these new features into tests that now use more points that were symbolically computed

✍️ Documentation

• updated README and examples with new Cython function and more extensive comparisons

📦 Packaging

• added a setup.py for the Cython compilation

Closes #4