More closely couple the API's loading bar with the calculations it's actually performing

[anth-volk]

At the moment, on the front end, we just make a guess as to how long the run should take. If technically feasible and not a heavy lift (and I'm not sure either of those are true), it'd be great to more closely couple the loading bar to the actual calculations we're doing. If we could evaluate the total queue size at some optimal point, then just check how many we've done, maybe this would be possible? And we use a front-end polling structure every ~1s, so all we'd have to do is return a percentage calculated from there.