Mark simulator as a cython package in setup.py

[bgmeulem]

Hiya

This PR proposes to mark pyloric.simulator.pyx as a Cython extension in seutp.py. This helps out some package managers.

I had some issues properly installing this package. I use pixi as a environment manager and front-end to build packages. This did not automatically build the Cython extension. In addition, running python setup.py build_ext also does not build it, since it's not marked as an extension in setup. I didn't check whether or not this issue exists when using pip directly, as advertised in the readme. However, I suspect it's good practice anyways to mark Cython packages as such in the setup.py. I noticed they need the numpy headers to compile the .c file, hence the additional include_dirs argument in setup

[bgmeulem]

After some digging around, I think I understand some additional issues I'm encountering building the Cython module (apart from what's mentioned above i.e. declaring them in setup.py).
I encountered the following additional errors upon importing pyloric.simulator

ImportError: numpy.core.multiarray failed to import (auto-generated because you didn't call 'numpy.import_array()' after cimporting numpy; use '<void>numpy._import_array' to disable if you are certain you don't need it)

Can be fixed by simply adding the requested line in simulator.pyx (see ce405c6)

ValueError: numpy.ndarray size changed, may indicate binary incompatibility. Expected 96 from C header, got 88 from PyObject`

Seems like numpy changed the C headers to expect ndarrays of size 80 prior to numpy==1.20, size 88 after numpy==1.20 and size 96 god knows when, but likely numpy==2.0. So if you build the thing with a numpy version that is too recent (not unlikely in isolated build environments), or likewise have a numpy version that is rather old in your runtime environment, you will get this mismatch.

Solutions:

• Version track pyloric to match numpy versions
• Building the package within your run environment using --no-build-isolation. In this case, the build dependencies in pyproject.toml are ignored, but the package is built using whichever numpy version exists in your environment when you invoke the build.