[clang] [CUDA][Win32] Add `fma(long double,..)` to math forward declares. (PR #73756)

[Joachim Meyer via cfe-commits]

fodinabor wrote:

On the AdaptiveCpp side of things, we are working around this issue by having our own wrapper wrapper header file: https://github.com/AdaptiveCpp/AdaptiveCpp/blob/develop/include/hipSYCL/std/hiplike/cmath

So the change on the LLVM side is seemingly still outstanding.

https://github.com/llvm/llvm-project/pull/73756