Fix short transcripts

[AnnaLazarEBI]

FIlter for 1 codon translations (3bp).

Tested.

Original gtf: /hps/nobackup/flicek/ensembl/genebuild/lazar/bbr/zymoseptoria_tritici_pangenome/GCA_017766825.1/annotation_output/initial_region_gtfs/4.rs1.re2952612.busco_copy.gtf

Before filtering:
/hps/nobackup/flicek/ensembl/genebuild/lazar/dev/validation/old/output.gtf

After filtering:
/hps/nobackup/flicek/ensembl/genebuild/lazar/dev/validation/output.gtf

[ens-ftricomi]

we compute the translation before so I think we want to exclude these ?

[AnnaLazarEBI]

Issue 3bp CDS survived original filter.
For the 3bp BUSCO transcript to survive this:

1. ‎$transcript->translation must be ‎undef at that time

→ it goes through the ‎unless ($translation) branch and is kept.This happens if:

▫ you were in the “final transcript in file” case where ‎compute_translation was commented out when the transcript was first built, or

▫ it’s a transcript produced later (e.g. by joining / UTR processing) that hasn’t had ‎translation set yet.

2. Or it had a translation with ‎length > 1 aa at that moment, and only later (after you call ‎compute_translation in the ‎"Computing translations" loop) did the CDS collapse to 3bp.

We now filter out 1‑codon transcripts immediately after grouping and optional joining, instead of only relying on later CDS‑length criteria. The change walks over ‎joined_transcripts, inspects any existing ‎translation, and removes transcripts whose translation length is ≤1 amino acid, logging what was dropped. This prevents obviously spurious 3bp coding models (like BUSCO artefacts) from propagating into downstream steps and the final GTF, while leaving non‑coding or untranslated models untouched for later processing.

[AnnaLazarEBI]

These are filtered:
DS995906.1 anno exon 540775 543237 . - . gene_id "anno_4596"; transcript_id "anno_4596"; exon_number "1"; DS995906.1 anno transcript 1549067 1549069 . - . gene_id "anno_11407"; transcript_id "anno_11407"; biotype "busco"; translation_coords "1549067:1549069:1:1549067:1549069:3"; DS995906.1 anno exon 1549067 1549069 . - . gene_id "anno_11407"; transcript_id "anno_11407"; exon_number "1";

The first line is not expected.
original slice: /hps/nobackup/flicek/ensembl/genebuild/lazar/bbr/reannotation/batch2/GCA_000001985.1/annotation_output/slect_script_test/DS995906.1.rs1.re1818470.busco.gtf
filtered slice: /hps/nobackup/flicek/ensembl/genebuild/lazar/bbr/reannotation/batch2/GCA_000001985.1/annotation_output/slect_script_test/DS995906.1.rs1.re1818470.busco_sel.gtf