This is in preparation for a  ubsample script to call these
python functions directly, as opposed to scripts called from the
command line. In time, this script will become `augur subsample`.

The work done by the functions is unchanged, thus they still don't
return any information rather they write information to disk. Future
work will change this so that information can flow back to the calling
program without needing to be written to disk.

This implements a new script to encapsulate the subsampling logic formerly encoded in snakemake rules. This is in preparation for moving this script to the augur repo where it will become `augur subsample`. (We have chosen to develop this in the ncov repo for simplicity.)

The script currently uses the same approach as the former snakemake rules, however python functions are called rather than scripts / augur commands. Briefly, the steps are:

1. A subsampling scheme is provided, parsed, validated, and turned into a simple graph to indicate which samples rely on other samples having been computed (i.e. which are needed for priorities)
2. Each sample is computed by calling the run function of augur filter
3. If priorities need to be calculated for a sample to be computed, this is achieved by calling functions from the two existing scripts.
4. The set of sequences to include in each sample is combined, and outputs written.