microsoft/spirv_to_dxil: Push constant lowering fixes
Found two issues related to the push constant lowering when testing on DOOM. I'm pretty confident the first fix is legitimate, but I don't remember why we were retrieving the min/max push_constant slots instead of just the highest slot. To me, it looks like we'd end up with an invalid size if all those accesses have nir_intrinsic_base() != 0, but I'm not sure. @lfrb, do you remember why you did that?
Edited by Boris Brezillon