Support u16 and f32 pixel image layout. #105
Description
This is the reason why loading some tiff images fail (#98) but this isn't limited to tiff.
At the moment we ignore u16 and f32 varients of pixel formatting, we need to allow up to f32 in the codebase and then convert decoded images to RGBA8 before uploading to the GPU as par egui's requirement.
This is my understanding of the situation, I'm still trying to understand image colour formats.
FIR seems to offer colorspace converting so maybe we can make use of it: https://github.com/Cykooz/fast_image_resize
However this seems like a more convenient library for us: https://github.com/kbalt/ezk-image