[Bug libstdc++/125624] C++23: support of std::float16_t for std::uniform_real_distribution

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125624

--- Comment #3 from Jonathan Wakely <redi at gcc dot gnu.org> ---
We don't support them, they just happen to compile and produce *some* answer,
without any explicit effort from us or any intentional support. And for
float16_t it also produces *some* answer, which happens to be always NaN.

But the new implementation of std::generate_canonical in GCC 16 is known to not
work at all for float16_t.

Your concepts which check that the distribution can be constructed do *not*
mean the distribution works with those types.