Make use of array jobs on cluster

[alexdewar]

@davidorme reports that submitting individual jobs is slower. In any case, grouping some simulations together seems likely to be a bit faster in general

It may be that we can achieve this with Snakemake's job grouping feature: https://snakemake.readthedocs.io/en/stable/executing/grouping.html