[ub:expr.shift.neg.and.width] Use std::uint32_t instead of int in example

[eisenwave]

Firstly, I find it unnatural to use int and then have to assert various properties to get 32-bit type.

Secondly, the example is incorrect (well, not really, it's still true that a 32-bit shift is UB) because sizeof(int) == 4 && CHAR_BIT == 8 is not a sufficient assertion. It's possible that int is a 16-bit type, with two 8-bit padding bytes and two 8-bit value bytes.

Alternatively, we could also use numeric_limits<int>::digits in the assertion or just put the assumption about the width of int into prose.

[notadragon]

We should have the #include <cstdint>. I would also prefer that we roll this into https://isocpp.org/files/papers/D4284R0.pdf for discoverability, but otherwise this is fine.