Detangling input/output variables

[nichollsh]

This PR makes a significant step towards reworking how PROTEUS handles variables. With this version, output variables are stored as a Pandas DataFrame (the helpfile, HF). However, this HF is no longer passed around to as many functions as before, and is never modified by them directly. Calls to submodules (e.g. SPIDER, JANUS) only return the variables which they have calculated, which are then used to update the current working row of the HF. Once an iteration of the model is completed, the new row is committed to the HF and the whole DataFrame is saved to a CSV file.

Similarly, the COUPLER_options dictionary now exclusively stores model input variables. These are never modified at runtime and are not returned from any functions other than when the configuration file is first read-in.

The model runs and reproduces the previous results, which is expected given that these modifications are not intended to modify any physics. However, the output columns of the helpfile have different names in some cases, so this may break analysis or post-processing scripts written externally to the code itself.

Closes #95.

Other changes:

• Rework plot_atmosphere function
• Rework plot_fluxes functions
• Can now change plotting output format (pdf, png, etc) via the configuration file
• Calls to plotting scripts from the terminal have a consistent command line interface
• More documentation inside the code
• Removed some redundant library imports
• Modifications to AGNI wrapper
• Updated vscode settings for workspace
• Optimisations to surface_gases
• Moved default configuration (previously called init_coupler.cfg) to input/default.cfg
• Some efforts towards a consistent code style, with line lengths of <=92 chars and docstrings.

[nichollsh]

Next step after this will be to merge my resume branch, which includes functionality for resuming the simulation from the disk. https://github.com/FormingWorlds/PROTEUS/tree/resume

[nichollsh]

Thanks Tim! I agree it's hard to tell through all of the changes but it does seem to reproduce the same results as before for the pure-steam Earth example case.

We should think about adding some tests.