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?