CCAPS IO - slow python startup times are a bottle neck

[JOThurgood]

IO currently outputs simple .dat files and immediately calls a python script to plot them. This can lead to many calls to python (once for each dump). The python plotting script is substantially slower than the rest of the code.

Ideally

• output sequentially numbered dat files at each dump
• create a single post-processing script to do the simple visualisations at the end (or, perhaps at the start and have it wait on new dump files) to avoid multiple startup and exits of python
• this also has benefit of paving the way towards keeping dat files for more complicated post-processing / analysis

[JOThurgood]

Not just start up times bottlenecking - once getting to about 512x512 a lot of the plots take very long to do and the python script seems to hog a lot of memory (much than the simulation itself) during the process.