Optimization have improved over time and now LLVM manages to optimize more in-place-collect noop-iterators to O(1) functions. This updates the codegen test to match.