making room for all conceivable characters within a language, and even accommodating multiple languages
reads like snakeoil… Why not design the algorithm to work on arbitrary blocks of bytes so you don't need to care about „characters“, just like any real-world encryption algorithm?
Writing the roman numeral IIII instead of IV is not uncommon on clocks: https://www.deutsches-uhrenmuseum.de/en/museum/knowledge/clock-facts/clock-dials-with-4-iiii-or-iv.html