Support for reading higher bit-depth images

[luzlab]

We're developing machine vision solutions for in-process monitoring of manufacturing processes, including collecting wide-field absorbance maps. The camera is 16-bit and optimized for collecting low-noise quantitative data suitable for feeding deconvolution/superresolution analysis pipelines.

We currently use OpenCV to process the images, and generate 16-bit unsigned single channel and 32-bit floating point single channel data. Both formats are supported by OpenCV.

The current implementation of ImageSchema creates only 8bit image dataframes.