feat(oriongeno): add Nextflow and WDL annotation modules

[alejandrogzi]

Summary

Adds OrionGeno annotation modules for Nextflow and WDL. Both use ghcr.io/hillerlab/oriongeno:latest.

• ORIONGENO_PREDICT and the WDL predict task run one GPU. Output is GTF by default, or GFF3 when meta.oriongeno_format / format is gff3.
• ORIONGENO_MULTI and the WDL multi task run oriongeno multi and write GTF. Nextflow passes the GPUs in CUDA_VISIBLE_DEVICES. process_gpu requests one GPU under -profile gpu; set accelerator on ORIONGENO_MULTI to use more. WDL takes an explicit devices list.
• lineage selects the checkpoint. The default is mammals. Species name is only species conditioning. A checkpoint directory overrides the lineage.
• Repeat annotation is optional. Inference requires an NVIDIA GPU. Bundled OrionGeno is non-commercial.

Test plan

• Nextflow 24.10.5 -stub-run writes the annotation and, when requested, the repeat file
• Rendered commands carry the checkpoint, species, format, and GPU list
• An unknown lineage fails before the task starts
• ORIONGENO_MULTI rejects meta.oriongeno_format = 'gff3'
• A custom checkpoint path is passed through
• miniwdl parses both WDL files
• Annotation of a real genome on a GPU