PyJulia incompatibility on Linux

[blakeaw]

Describe the bug
PyJulia only runs correctly out of the box on Linux systems when the installed Python is built with --enable-shared. However, installing Python in a conda environment on Linux yields a statically linked Python that PyJulia doesn't support.

Python Environment (please complete the following information):

• Environment (e.g., conda): conda inside a Docker container with Ubuntu 16.04 base image
• Python version: 3.10.11
• GAlibrate version: 0.7.1

Additional context

The troubleshooting guide in the PyJulia docs, https://pyjulia.readthedocs.io/en/latest/troubleshooting.html#your-python-interpreter-is-statically-linked-to-libpython, includes several workarounds.

Potential Fix

I think for GAlibrate, the Turn off compilation cache workaround probably makes the most sense:

from julia.api import Julia
jl = Julia(compiled_modules=False)

I don't think there should be a large performance penalty since the GAlibrate functions in Julia need to be JIT compiled anyways when first called. Although I guess it could add some latency to PyCall.

The run_gao_julia.py module could be patched to check the OS (using sys.platform) and for linux it can apply the above Turn off compilation cache workaround. Something like:

from sys import platform

if platform == 'linux':
    from julia.api import Julia
    jl = Julia(compiled_modules=False)

[blakeaw]

Fix (workaround noted above) implemented in PR #15