Return predictions from a still image with web parity

[albullington]

This issue is a pre-requisite to completing #807 in the iNatRN repo, and it builds on #8, which exposes a function to allow predictions from still images the way we needed in Seek.

The goal here is to get parity with the way we show predictions on the web. Instead of returning a single branch of the tree of life, we should return results from multiple branches, plus common ancestors as needed.

Patrick put together a working doc of how we score and display images on the web to use as a guide.

We should skip the geomodel steps for now and address those in a separate, future issue.

[kueda]

@jtklein has done most (all?) of this in vision-camera-plugin-inatvision, but it still needs to be merged into main, and the RN app needs to be updated accordingly.