It would no longer be UTF-8; it would be UTF-32. UTF-8 is an encoding scheme, meaning that it is a specification for exactly how text is encoded as bits.
You can certainly use UTF-32 to represent all valid unicode, but you can only do that within the bounds of a single program; once you need to read or write data to or from an external source (say, the file system, or over a network), you’d need to use the same encoding that the other software uses, which is usually UTF-8 (and almost never UTF-32).
Oh, Nim is possibly even a better example because it is “transpiled” rather than compiled, meaning the compiler actually generates C or C++ code. You can then compile that with whatever compiler you want. However, I don’t know of any major projects in Nim to compare against ones in C, C++, Rust, etc.
Edit: and Zig should be extremely efficient as well.