Updates to min_model_error, to bring it up to date with changes in eric_devel etc.

[aliceramsden]

This branch can be merged into min_model_error when everyone else is ready to start turning min_model_error into the new 'main' branch of the repo.

Updates on this branch include:

• A fix for reading in the correct prior fluxes, when creating the posterior country fluxes and saving everything after the inversion. The prior fluxes are now read directly from the merged data object, and the correct monthly/annual flux is sliced from the full flux object. This includes taking an average flux across a range of months, if the inversion is across mulitple months.

• A try/except loop which drops sites from the inversion if the data merge process doesn't work for that site (which normally happens if there's no obs).

• Adding in a print out of the number and % of obs that are removed by each filter, at each site.

  • Fixes for saving and reading in the merged data object, including modifying the site variable (and associated heights etc.) if these aren't found in the merged data object.

• Some minor bug fixes, including some in the basis function creation process and some variable names.

[brendan-m-murphy]

Hi Alice, thanks for this, I'll work on reviewing it and merging into min_model_error tomorrow.