Consider restricting to one preferred layer per multi-target score set

[sallybg]

Currently, the mapper allows each target in a multi-target score set to have its own preferred layer (protein or genomic), depending on the available HGVS strings for that target. However, I'm not sure it would ever come up to have an experimental assay where some variants are assayed as nucleotide variants and others are assayed as amino acid variants. This is low-priority since it currently works as-is, and may become irrelevant if we end up expanding mapper output to produce c., p., and g. coordinates where possible.