Support 16-bit qtables for 8-bit data

[kornelski]

Unfortunately 16-bit quantization tables in 8-bit files are a de-facto standard: libjpeg-turbo/libjpeg-turbo#90

This image:

trips the assertion in:

jpeg-rust/src/jpeg/jpeg.rs

Line 231 in 24655ed

assert!(precision == 0);

[martinhath]

Cool!

I have added support for 16-bit tables, but I am not sure if it works, as it seems the image in your post is using progressive DTC, which currently is not supported (which I guess is also why there is a 16-bit quant table here, as both 8-bit and 16-bit are allowed in progressive DTC)