[clang] [clang][bytecode] Yet another __builtin_constant_p implementation (PR #130143)

[Nikolas Klauser via cfe-commits]

philnik777 wrote:

I think this makes sense to me. However, it's entirely unclear to me how you prevent modifications for cases like
```c++
bool test(int& i) {
  return __builtin_constant_p(i = 5); // How does this fail?
}
```

> Okay, this fixes 112 libc++ tests.

BTW, this is awesome! The rate at which the libc++ test failures are going down makes me think we might be able to switch to the new interpreter in a year or so. I'm really interested how much of a test speedup we can get from this.

https://github.com/llvm/llvm-project/pull/130143