[Bug middle-end/123869] Negation of _Float16 should just flip the bit and not promote to _Float32

[pinskia at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123869

--- Comment #9 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Just tested a few more things just in case.
```
_Float16 f1(_Float16 a, _Float16 b)
{
  return __builtin_copysignf16(a, b);
}
```
Works as expected on aarch64 and x86, expanding using bitwise currently (good).

```
_Float16 f1(_Float16 a)
{
  return __builtin_copysignf16(a, -1.0f16);
}
```
Fails without my patch but works also with my patch as the above is done as:
```
_Float16 f1(_Float16 a)
{
  return -__builtin_fabsf16(a);
}
```

So don't need to change my patch for copysign after all.