Fix a bug in buffer_buffer_shader_timings.ln

[dfellis]

I realized the mistake after I tried to run the earlier version on MacOS and it crashed the machine (!!!) and then realized what I did wrong when I actually printed the buffer once it was in RAM and saw the garbage that was produced, particularly with the first few, which didn't change the buffer at all.

I was accidentally setting the yMax to 0, which Vulkan and DirectX decide this means "don't run the shader at all" but Metal decides "run the shader on ALL of the GPU memory address space (afaict) until it messes with something it shouldn't and blows up the machine.

This fixes the bug here explicitly, but I need to decide what to do if you call gFor with invalid arguments ("correct" the 0 to 1? panic! the program? make gFor fallible and return an error type?)