future feature: how to deal with multiple thumbnails?

ImageInput and ImageBuf() have `get_thumbnail()` methods. But they assume that there are 0 or 1 thumbnails per subimage. But some raw camera formats (and libraw) supports multiple reduced-resolution versions of the main image. We don't particularly have a way for an app to find out that there are multiple thumbnails, what their resolutions are, or to pick a specific one.

Maybe one possible route for the future is to extend the get_thumbnail() method to take a thumbnail index, to select which one is chosen.

Note that this would be a breaking change to the ImageInput API, so could not be backported to existing releases. (And in main, it would require making a v3_2 namespace version of ImageInput.)

Start with the ImageInput and ImageBuf get_thumbnail() methods mentioned in the issue, then review how raw camera formats and libraw expose multiple reduced-resolution images. Define how callers would discover thumbnail counts and resolutions, select an index, and preserve the required v3_2 ImageInput namespace compatibility. Done requires an agreed API design and tests covering multiple thumbnails.

Written by the indexing model from the issue text.