[alejandrogzi]

Summary

Adds the standalone OrionGeno pipeline, matching the Tiberius and ANNEVO workflow layer:

• workflows/oriongeno requires --genome and --lineage. --checkpoint overrides the lineage checkpoint. --species_name is species conditioning only.
• subworkflows/oriongeno scatters none / chromosome / weighted and runs ORIONGENO_PREDICT (one GPU per piece). ORIONGENO_MULTI stays a separate module.
• Chromosome and weighted modes gather annotations in FASTA order and renumber gene ids (g1, g1.t1, …) and repeat ids (r1, r2, …), because each predict job restarts at g1 / r1.
• --oriongeno_format is gtf (default), gff, or gff3. Repeats are emitted only with --oriongeno_output_repeat.

nextflow run workflows/oriongeno/main.nf \
  -c hillerlab.config -profile gpu \
  --genome genome.fa \
  --lineage mammals

Test plan

• Nextflow 24.10.5: none, chromosome, and weighted produce the same GTF, in FASTA order rather than chromosome-name order
• GFF3 child ids (exon, CDS) and repeat gathering follow that same order
• Unknown lineage, bad format, and a chromosome scatter above 1000 sequences fail before predict jobs
• Real GPU annotation (not run here; no NVIDIA driver)