Snakemake improvements mk2

A grab-bag of things I've noticed. ## Subsampling of genome-focused filtering This wasn't necessary for B3.13 / D1.1 as we explicitly wanted to analyse everything but for other builds it's desirable. With the current DAG it can be achieved by adding `--subsample-max-sequences` to `rule filter_metadata_by_subtype`. The segment-focused build downsamples in `rule filter` and we should see how to unify these. ## Hardcoded subtypes for genome-focused builds The `genome.smk` has lots of constraints like 👇 , but now that we're using two snakefile entrypoints I think we can entirely drop them. Needs testing! This is a massive limitation as it requires snakemake editing to make a new genome build. If we need them for some reason we should be able to dynamically generate them from the config. The `auspice_name_to_wildcard_name` function needs to be similarly updated. ``` wildcard_constraints: subtype = 'h5n1-cattle-outbreak|h5n1-d1.1', ``` ## Colour assignment errors if too many demes We whould make this work by repeating colours. It's frustrating to have the entire workflow error because there are too many demes in the analysis.