Support inputs with higher image depth

Currently Xpano converts inputs to 8bit images.

Stitching 16bit images with OpenCV was tested here: https://github.com/dankamongmen/stitcher16

todo: check how the current [stitcher pipeline](https://docs.opencv.org/4.x/d1/d46/group__stitching.html) handles 16 bit inputs.

Start by reviewing the current stitcher pipeline and OpenCV's stitching documentation, then compare its 16-bit behavior with the linked stitcher16 project. Trace where Xpano converts inputs to 8-bit and determine whether higher-depth images can pass through the pipeline. Done means supported higher-depth inputs are preserved through stitching without the current forced 8-bit conversion.

Written by the indexing model from the issue text.