[llvm] AMDGPU: Improve codegen for intrinsic llvm.fptrunc.round (PR #104486)

[Changpeng Fang via llvm-commits]

changpeng wrote:

> Do you have an example where codegen is improved?
Same code generated for round.upward ad round.downward. Add two additional cases round.tonearst and round.towardzero.
 
> How about calling it `FPTRUNC_ROUND_F16_F32` to match `V_CVT_F16_F32`?
Will do as suggested. Thanks.

https://github.com/llvm/llvm-project/pull/104486