Adding tabix'd GFF3 support to NearestGene

[jamie-m-a]

Summary

This PR allows NearestGene to retrieve gene locations from a tabix-indexed GFF3 file, enabling offline use:
--plugin NearestGene,gff3=/path/to/genes.gff.gz

It also adds:

• parent_only=1 to use gene-parent feature types supported by VEP’s GFF transcript construction, excluding artifact and readthrough-only genes
• regulatory=0 to avoid automatically enabling regulatory annotation
• Distance and forward-strand direction when using both_directions=1

Existing database-backed behaviour is unchanged.

Can test it by...

Using the release 116 GRCh38 GFF and FASTA, create an input VCF:

printf '%s\n' \
  '##fileformat=VCFv4.2' \
  '#CHROM	POS	ID	REF	ALT	QUAL	FILTER	INFO' \
  '1	157422	.	G	C	.	.	.' > test.vcf

Check the plugin compiles:
perl -c NearestGene.pm

Run VEP:

vep \
  -i test.vcf --vcf --offline \
  --fasta Homo_sapiens.GRCh38.dna.toplevel.fa.gz \
  --gff Homo_sapiens.GRCh38.116.chr.gff.gz \
  --plugin NearestGene,gff3=Homo_sapiens.GRCh38.116.chr.gff.gz,regulatory=0,both_directions=1,parent_only=1 \
  --force_overwrite -o test.out.vcf

The NearestGene result should contain:
ENSG00000269981:19457:upstream&ENSG00000279928:25274:downstream

The distances correspond to the GFF coordinates:
ENSG00000269981 ends at 137965
157422 - 137965 = 19457

ENSG00000279928 starts at 182696
182696 - 157422 = 25274

Note that running the same variant using the database will not fetch nearest genes (the plugin only works for intergenic variants), and these are annotated as downstream/upstream in the DB as they around/in the transcripts of a lnRNA. This is a limitation of the transcript model building of GFF3 input, it has fallen out of sync and will be updated in future to account for more gene biotypes.