Add screenshot diff to failed device tests in CI

[TychoVrahe]

If device test fails in CI due to screens being different, it would be nice to highlight in report where the screens differ, since sometimes the differences might be too small for human eye.

ImageMagick compare tool is one option, see example output in this comment https://github.com/trezor/trezor-firmware/pull/2471#issuecomment-1243346467