[clang] [llvm] [HLSL] Implement the f16tof32() intrinsic (PR #165860)

[Steven Perron via llvm-commits]

https://github.com/s-perron commented:

This looks good from an SPIR-V perspective. However, add a comment explain why you don't manually unpack the inputs. It would be fewer spir-v instructions.

https://github.com/llvm/llvm-project/pull/165860