Improve species percolation logic to better deal with the extinct case

[davidebbo]

For extinct species, we sometimes have good images, like 'Life reconstruction', and sometimes unpleasant ones, like a few fossilized bones/teeth.

Today, the 'bad' ones can end up percolating up, because we only rely on image quality in picProcess.py (and we don't know much about quality).

But if we were to add popularity to the percolation calculation, we probably would end up with better images, as more popular species are more likely to have nicer images.

[hyanwong]

Can we use popularity as a straight replacement for image quality somehow, e.g. by normalising it to within the same range as the quality scores?

I can't remember exactly how we use the quality scores in the picProcess algorithm, and how much is hard-coded to the 10000 - 50000 range we use for quality scores. @jrosindell will know, as he wrote that code.

[davidebbo]

Copying my slack comment: Yes. But one obvious gotcha: we'd end up mostly percolating extant species, as they're likely far more popular than extinct (e.g. no ancient bear can compete with polar bears or modern pandas!). I suppose we could boost up extinct over extant.

@jrosindell wrote: I see, good point - probably we should percolate both extant and extinct species separately based on quality / popularity and only use extinct signposts if no extant species exists. I prefer this to augmenting the popularity which will feel a bit made up and hard to get right.