Arguably this is somewhat of a kludge(?), but one option (which would involve less coding than making the engine work with different encodings) would be to use

Arguably this is somewhat of a kludge(?), but one option (which would involve less coding than making the engine work with different encodings) would be to use readers/writers to support other formats, and keep using UTF-8 internally. UTF-16BE/LE and UTF-32, plus Windows CPs, are all really easy to do like this — probably others too. (Note that depending on the range of codepoints used in files using these encodings, they may take more memory in UTF-8 form. This usually isn't an issue unless one is handling large files of Kanji)