Split input fasta file, smartly

[chasemc]

This might only be important with quite large metagenomes and even then probably will still not be the slow step even if uneven. Currently, when running in parallel, the input contigs are naively split/chunked into x-number of files. We had discussed before that it might be good to split contigs a bit more evenly into files based on contain length so that computation will be a bit more even later.

Relevant code: https://github.com/KwanLab/Autometa/blob/f372b8234766cc5e413a6c4671f78968ae682043/modules/local/seqkit_split.nf#L31-L36

https://github.com/KwanLab/Autometa/blob/f372b8234766cc5e413a6c4671f78968ae682043/conf/modules.config#L78-L82