Release builds of simple-raytracer are now benchmarked too. Release builds are slower but should still be faster than the LLVM backend.

Is the "faster, slower" here referring to compile times?