P3008R6: Atomic floating-point min/max

**Link:** https://wg21.link/P3008R6

Feature-test macro (increased):
```c++
#define __cpp_lib_atomic_min_max 202506L // freestanding, also in <atomic>
```

Note that `<cmath>`/`<math.h>` are also touched. This part is related to #105174, #139885, and #148133.