Reconsider start and stop in the IO methods

[camilogarciabotero]

The ORF struct is normally defined with a location field that is of type UnitRange{Int64}. This has been used with the default step (i.e., 1) argument. So even if the strand field of ORF is - the start will always be determined by the "positive" strand range.

This is not an issue for the get_orfs_* methods since they use the following treatment:

Base.getindex(sequence::NucleicSeqOrView{A}, orf::ORF) where {A} = orf.strand == '+' ? (@view sequence[orf.location]) : reverse_complement(@view sequence[orf.location])

The inverted range is, for instance, how negative stranded ORF are displayed in PHANOTATE outputs (c.f source code).

The things to reconsider are:

• Are the other ORF applications using this convention as well?
• Would revamping this bring some benefits to the performance?
• The write methods should at least advertise this. However, judging by the previous test with IGV it is found to have only positive ranges at start and stop.