WIP: port pylandau to use Numba

[leloup314]

This PR ports the pylandau package from using Cython-extensions, generated from cpp source code, to architecture-specific, pre-compiled bytecode from pure Python scripts using Numbas Ahead-Of-Time (AOT) compilation capabilities.

All functionality has been ported already, just the package structure needs to be adapted and Numbas AOT API has to be integrated into the package setup.

Performance stays the same:

Cython-based (current) version:

In [8]: a = np.arange(0, 100, 0.01)

In [9]: %timeit pylandau.landau(a)
1.03 ms ± 19.6 µs per loop (mean ± std. dev. of 7 runs, 1,000 loops each)

In [10]: %timeit pylandau.langau(a)
19.1 ms ± 15.3 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

Numba-based version:

In [13]: a = np.arange(0, 100, 0.01)

In [14]: %timeit pylandau.landau(a)
1.02 ms ± 2.17 µs per loop (mean ± std. dev. of 7 runs, 1,000 loops each)

In [15]: %timeit pylandau.langau(a)
17.6 ms ± 24.4 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

[leloup314]

Test are not expected to succeed due package structure / setup not working yet.

[leloup314]

Related to https://github.com/SiLab-Bonn/beam_telescope_analysis/issues/20

[leloup314]

@cbespin let's have a chat on how to integrate this

[cbespin]

Related to https://github.com/SiLab-Bonn/beam_telescope_analysis/issues/20

At least this is solve with today's release (coming to pypi tomorrow).

I am always down to get away from cython, we can discuss the details. Changes look good so far

[leloup314]

Move development to new branch in this repo