Make Plotting optional and extendible

[jtravs]

Use PythonPlot instead of PyPlot, to try and solve mac crash issues
This is WIP because it seems PythonPlot keeps Python 0-based indexing rather than converting to 1-based.

[chrisbrahms]

Does this actually fix the crash issues?

[jtravs]

Does this actually fix the crash issues?

Not sure yet, but given that PyPlot is essentially unmaintained in favour of PythonPlot, this makes sense anyway. I will test the crash issues this evening. I think I have this working now.

[chrisbrahms]

That’s reasonable. Thanks for sorting this. I suppose we should really have a test script to make sure all the plotting functions run without error

[jtravs]

This just got a lot harder than I expected. Anywhere there is indexing this messes up.

[jtravs]

I need to leave this for now. For my future reference, this currently errors on a time_1d plot for a multimode simulation, on line Plotting.jl:637. Basically, we call Maths.fwhm on the data from the plot itself.

[jtravs]

I believe this is complete now. AFAICT all Plotting functions work with both single and multimode output. I didn;t check all of the examples. Early indications are that this doesn't fix the crashing on mac.

[chrisbrahms]

This PR seems to be required now. It fixes the issue with installing conda from scratch on Ubuntu, which has existed for some people since at least March (see e.g. https://discourse.julialang.org/t/conda-not-installing-matplotlib-for-pyplot/96813/3). This is now getting more important as it prevents the new release of Luna to be auto-merged.

Couple of small issues to be fixed still, working on those now.

[jtravs]

@chrisbrahms this has been significantly overhauled. It now uses Julia extensions to enable optional code loading. This has two big benefits:

1. The code can be run without requiring Python or other plotting packages, e.g. on HPC systems
2. We can try different plotting backends

Currently, I have implemented PythonPlot, PyPlot and Makie. They all seem to work fine. I would suggest that we recommend PythonPlot as it is the supported matplotlib interface. Personally I am using Makie more as it crashes less on mac.

Still to do:

• Remove gratuitously duplicated code in the backend files
• Test much more
• Check examples and documentation for consistency