[libclc] libclc: Replace nextafter implementation (PR #186082)

[Matt Arsenault via cfe-commits]

arsenm wrote:

> I think we don't want to use any floating point operation in the nextafter implementation.

This gives much better codegen. At worst this should check around if DAZ is enabled for the type.  I'm assuming this needs a canonicalize placed somewhere 


https://github.com/llvm/llvm-project/pull/186082