Add support for this builtin:
double builtin_truncf128_round_to_odd(float128)

Differential Revision: https://reviews.llvm.org/D48483

llvm-svn: 336595