Split UniProt Mapping Jobs into per-target rather than per-score set

[bencap]

At the moment, UniProt mapping jobs operate on a per-score set basis, with all targets in the score set attempting to be mapped and then polled. It would be easier to reason with if we split this out into a per-target gene operation, since all targets should have the same workflow. To do this, we'll likely need to add some sort of setup job that creates mapping and polling job runs for each uniprot mapping operation we need to complete.