[libclc] [libclc] Move smoothstep to CLC and optimize its codegen (PR #123183)

[Fraser Cormack via cfe-commits]

frasercrmck wrote:

Here's an example of the difference in LLVM IR for `smoothstep(double16, double16, double16)`: https://godbolt.org/z/P9sM3Wjjn



https://github.com/llvm/llvm-project/pull/123183