[flang] Quadmath 128 bit floating point intrinsics

This update allows constant folding for many 128 bit floating point intrinsics
through the library quadmath, which is only available on some platforms.

Differential Revision: https://reviews.llvm.org/D156435