That's one of those things that, if you ask me, should fall outside the scope of a character set. As I see it, they are just the usual numbers, and restyling, resizing and positioning them should be a concern of a "text" layer, above the "character" layer.

It should. But Unicode's mandate included incorporating common existing character sets, and ISO8859-1 had ² and ³, and it all when down from there, so that now you can write 𝘌 = 𝘮𝘤² which isn't the same as E = mc².

A number of early computer systems used teletypes for I/O, and had programming languages in which exponentiation was expressed by moving a half line up, and array subscripts by moving a half line down.

I seem to recall the designer of ascii stating that it was a mistake, and they should've gone with prefix coding instead (following-is-capital,a->A;following is number,<byte>->67 or something along those lines - basically escape codes for everything, I suppose). Instead we got ascii evolved to really, really, really big ascii (utf-32, and the variable encoding utf-8).

But yeah, having a "next-is-superscript"-token would probably be cleaner than individual code points for every little glyph that might be a superscript...

In that case, you're really going to love the blackletter block:

https://en.m.wikipedia.org/wiki/Blackletter#Unicode

Maybe in an ideal world, sure, but in this world, nobody is going to add those features to the text layer as long as there are perfectly good Unicode characters right there in the character layer. So we may as well use them.

Most styling markup formats (except markdown, irritatingly) have added those features to the "text" layer.