Don't use _Float16 on architectures that do not support it

[mcatanzaro]

_Float16 is only supported on a few architectures. Let's assume it's not supported unless we know otherwise.

References:

• https://bugs.webkit.org/show_bug.cgi?id=274086
• https://gcc.gnu.org/onlinedocs/gcc/Half-Precision.html
• https://clang.llvm.org/docs/LanguageExtensions.html#half-precision-floating-point