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

[via cfe-commits]

zyzheng wrote:

Hi, I also encountered this issue and wondering what is the status of this pull request?

I modify `__clang_cuda_math_forward_declares.h` file locally with the proposed changes and the compilation passed. So I think `_MSC_VER` guard is ok.

Here is may command line:
```
> clang++.exe --version
clang version 21.1.6
Target: x86_64-pc-windows-msvc
Thread model: posix
InstalledDir: D:\Program\clang+llvm-21.1.6-x86_64-pc-windows-msvc\bin
> clang++.exe -isystem "C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.8/include" -Wno-unknown-cuda-version --cuda-gpu-arch=sm_89 --cuda-path="C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.8" -march=native -std=c++20 -x cuda -c 1.cu -o 1.cu.obj
```
`1.cu` is an empty file and MSVC version is 14.50.35717.

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