Analyze differences in CRAFTY-DE results from June 2025 to Octobre 2025 + technical issue

I have run the same CRAFTY-DE data (ssp1a26_RunA_comp_25-04-07_AFTCodes_base2019-yr2100.nc) with a version of this script made before and after the merge with the "optimize" branch (commit de73bc1d).

The results look good and are very similar, with only minor differences, especially in the historical values, which are likely related to stochastic functions.

However, I am having troubles to run the full gridlist both on my computer (which crash everytime at 80%) and the OWL node (even using #SBATCH --mem=256G). We need to find a solution. I don't understand what the "optimize" branch added to create these complications.

A temporary solution now is to run the grid in two separate files and merge them.