intel/compiler: move predicated_break out of backend loop
What does this MR do and why?
intel/compiler: move predicated_break out of backend loop
This has no impact on the generated shaders, but does have a small
(positive) impact on the amount of time spent in shader compilation.