Consider using slurm job arrays

[JamesWrigley]

When reprocessing lots of runs with e.g. amore-proto reprocess all, we always loop over each run one at a time and launch a slurm job for each one. For a proposal with 323 runs this takes about 40 minutes, which is a bit annoying if you only want to reprocess cluster variables anyway.

An alternative would be to launch all the slurm jobs at once at the beginning using job arrays: https://slurm.schedmd.com/job_array.html
And then loop over the runs to process the non-cluster variables as usual.