This is a very neat solution. With different translation tables you can have support for other languages, and you can easily test to see if translation failed: if you're on a Korean version of Decker and tried to load a deck that uses Latin1 characters, those would translate to the invalid Unicode character. It's "yet another standard" but it's an internal standard only, so some deliberate incompatibility doesn't seem that painful.
Chinese is still out, of course. Hiragana+Katakana doesn't fit, but Hiragana would on its own and you could sometimes use a Katakana font. The entire Cyrillic subset of Unicode wouldn't fit, but the top 8 Cyrillic languages altogether only need 110 letters. Right-to-left languages would fit in the translation tables but have other bigger problems with rendering.