[compiler-rt] [compiler-rt][builtins]Fix complex division for aarch64 (PR #106664)

[Eli Friedman via llvm-commits]

efriedma-quic wrote:

The C standard only says "The behavior of signaling NaNs in Annex G is implementation-defined".  C++ says nothing at all.  So the question here is just, is there some compelling reason to copy gcc here?  I'm not sure I see it, unless there's some gcc documentation suggesting their behavior here is intentional.

https://github.com/llvm/llvm-project/pull/106664