Support for non-Latin characters

[robertknight]

ocrs should support models that can recognize non-Latin text.

Some likely sub-tasks:

• Eliminating the hard-coded alphabet from the recognition process. IIRC this was from an EasyOCR model that I used at one point.
• Since the alphabet is deeply linked with the model, it would be convenient if both could be shipped in one .rten file.
• It will be necessary to figure out how to support a larger set of class outputs without too much of a performance hit.

[DehaiWang]

will support Chinese character？

[robertknight]

The goal is to make this possible. There are a lot of details still to be figured out.