Unicode normaliser

Converts text between Unicode normalisation forms. Two strings can look identical, print identically and still not match, because the same accented character has more than one valid encoding.

An é can be one codepoint (U+00E9) or two (e followed by a combining acute accent), and both render identically. macOS historically stored filenames decomposed while almost everything else composes, which is why a filename copied from a Mac can fail to match the same name typed on Linux — a bug that looks like magic until you know about normalisation. NFC composes and is what you almost always want for storage and comparison. NFKC additionally folds compatibility characters, turning the fullwidth Ａ into A and the ligature fi into fi, which is right for a search index and wrong if you need to preserve exactly what someone typed.