Improve photo download estimate, handle under-estimate more gracefully #27
Description
Currently we estimate photos based on the first page of observations, which suffers from recency bias. We could instead
- Buffer that estimate a bit more with a heuristic, e.g. just double it
- Update it with each page (never decreasing it)
The main problem is that there's no photos endpoint for the iNat API so there's no real way to know how many photos there are in an observation query. We could also make the progress bar say that the max is just an estimate.
Furthermore, when we've exceeded the estimate and we're still downloading photos, we definitely shouldn't show negative values for estimated time to completion.