Please no. This experiment was tried back in the 70s, with a language called APL. You got a special terminal, including special keyboard, with all sorts of special symbols to do various operations. Every key on the keyboard got an extra symbol, so they weren't limited to well-known math symbols, and then they added "overstruck" operators as well...
It was interactive and interpreted, and was probable designed to displace BASIC. When I was in university (77-81), my school taught APL to the non-engineering, non-science majors (CS majors learned PL/1, other science and engineering folk learned Fortran.) It's big advantage seemed to be that "interpreted" bit - user could load a 'workspace" and have access to high-level functions written (by someone else) for a more specific field (rather the way that Python is used today.)
AFAIK, the extended character set was considered an abject failure. It reduced availability, added complexity, and made code nearly unreadable. (now, this is also about the same timeframe that "normal people" (as opposed to would-be clerical staff) were learning to type. So making the keyboard much more complicated probably didn't help.)
Despite the fact that nearly all modern keyboard/display setups are capable of dealing with much broader character sets, I don't think that this is an experiment that needs repeating
(Further, I've never heard that the lack of character being a problem that needs solving, with the possible exception of a relative minority of mathematically inclined folks that are really upset by the "confusion" they think is caused by using the "=" "equality symbol" for assignment...)
(Finally, Arduino is at the mercy of the C/C++ compiler and language that it uses. Using unicode symbols in Arduino would require that they first be used in C++...)