[builtins] Generate __multc3 for z/OS (#77554)

https://github.com/llvm/llvm-project/pull/68132

 ended up removing
__multc3 & __divtc3 from compiler-rt library builds that have
QUAD_PRECISION but not TF_MODE due to missing int128 support. I added support for QUAD_PRECISION to
use the native hex float long double representation.

---------

Co-authored-by: Alexander Richardson <mail@alexrichardson.me>