Yes, XML / HTML coding looks to me as a good idea.
Except that (if I’m right…), names in Kicad are supposed to be ASCII only, for portability. But maybe it’s only in the libraries ?..
Using the XML encoding would probably requires rewriting the parser code.
(But, AFAIK, there is already an XML oriented lib included in Kicad).
NOTE : Using a C style notation for the “really” special characters as \n, \t, and keeping the printable characters, like ‘|’ or ‘:’ as-is may be simpler, both for the developers and the end user.
Also, drawing the Unicode characters onscreen, or plotting, may not be currently supported, as I believe it uses a graphics line oriented character drawing routine and a custom font.
Lack of fonts is a display issue. You want the input representation to be future-proof. If a character can’t be displayed, then display something like a block there and you’ve lost nothing anyway. For parsing all you have to do is pass the string to an XML entity decoder and you get back the Unicode string.