Run sort command without pre-calculated minimizer index

[ivan-aksamentov]

Context: https://discussion.nextstrain.org/t/nextclade-sort-using-a-serverless-local-dataset-for-sorting/1777

Nextclade CLI sort command currently requires minimizer_index.json, either from dataset server or one that's provided using -m parameter.

The minimizer_index.json contains a mapping from dataset/ref names to the list of minimizers (hashes of ref sequence fragments).

In principle, Nextclade CLI could calculate this index if given a set of reference sequences. The code already exists - it already has to calculate minimizers of query sequences, so it might as well calculate minimizers for reference sequences.

We could add a parameter to the sort command (e.g. --input-ref, -r) to provide one or more fasta files with ref sequences, and in this case Nextclade CLI would not require either -m or fetching the index from a server. It would instead calculate minimizer index from the provided sequences and then immediately proceed to sorting. This could also be implemented as a separate command which only generates the minimizer index file - this way the produced index file could be reused (with sort -m), without making repeated minimizer index calculations on every run.

This should improve user experience for people who want to use sort command with their own reference sequences. Currently they have to go through building a custom minimizer_index.json which is a non-trivial task.