[Changelog]

I/O

The year dimension of simulation inputs is stacked as a single column on disk (to be compliant with the parquet specification requiring string column names). If the input data is empty, float data type is assumed and other index levels are filled with NaNs on disk (but discarded on read).
Simulation outputs are batched into a single parquet file by variable at the end of the simulation to improve copy/transfer time.
When resuming a simulation, all resumed outputs are truncated at the resume start year if possible. The "Approach" index level is also renamed on disk if transferring a LogisticalStart to UseStart. The result of this step is automatically cached in the .cache directory and can be reused by other simulations resuming with the same parameters.

Added a script to allow running multiple scenario simulations in parallel (parallel_scenarii.py)

Changed the color palette of plots
Changed area plots to line plots in the selected indicators section of the interactive notebook.