[clang] [llvm] [AMDGPU] Enable sin/cos to sincos folding for HIP (PR #181774)

[Juan Manuel Martinez Caamaño via cfe-commits]

https://github.com/jmmartinez commented:

I have one question, why is this not a problem for OpenCL? Why is `__ocml_sincos_f32` preserved and not eliminated by DCE?

https://github.com/llvm/llvm-project/pull/181774