- Replace blast2go-convert-decypher with split-blast-hits + sedding.
  This is more flexible.

- Our RefSeq Decypher db used to have Hit_def and/or Hit_id formatted as
  gi|{ncbi GI number}|ref|{refseq id}| description. Now it just has a
  refseq id and description. The BLAST2GO script was updated to reflect
  this.

by making it streaming instead of reading whole file to memory

Not used in pipeline, but needed to serve a request for a manual CLIME
run

Unused because we assume PhyML outputs a bifurcating tree

- Rename $parameters variable so it doesn't name clash
- Give more memory
- A couple more fixes

and:

- Use pytil instead of chicken_turtle_util
- Remove unnecessary @UnresolvedImport

Only use core genes with gap positions removed. Use SPR
search. Assume resulting tree is bifurcating.

Run one per species, no chunking

Run one per species, no chunking

Single-core, only generate groups

Tree generation failed (dlcpar_search uses a recursive algorithm which
exceeds max recursion depth), so we tell OrthoFinder to generate no more
than ortho groups.

Tree generation failed (dlcpar_search uses a recursive algorithm which
exceeds max recursion depth), so we tell OrthoFinder to generate no more
than ortho groups.

Segfaults on cluster, probably due to some installed packages not
working on cluster, only on midas server