Refactor GTF/GFF code to split out RefSeq/Ensembl

[davmlaw]

When I started, I broke things down into GTF vs GFF

I think it's probably a better split to write a general reader that can retrieve things - then split it up into RefSeq vs Ensembl as there is a lot of speciic code creeping in

Better to be expliit on what we're grabbing

[davmlaw]

step 1 should be write lots of unit tests for GTF/GFF to JSON

[davmlaw]

Also, Ensembl only works for GTF now, fails handling protein versions for GFF3. We should perhaps add explicit tests for this if we only support 1 as error message is bad