Further improve set of validation datasets for end-to-end testing

[boydgreenfield]

First:

• Twist + Illumina
• ARTIC v1 + Illumina
• RNA-seq (Illumina)
• ARTIC v2 + Illumina
• ARTIC v3 + Illumina
• High human known negatives (to test polyA tail mapping and related issues)

Later:

• ARTIC + ONT
• dRNA-seq (ONT)

We need truth data for all of these so we can do some golden output testing, i.e., hand inspected variants that we truly believe. This may or may not be doable for some of the RNA-seq data.