unify load_ubo_dxil and load_ubo_vec4
While developing the d3d12 driver, we added a load_ubo_dxil
that is effectively the same as load_ubo_vec4
, but our lowering-pass seems to be a bit more powerful, due to requirements in the OpenCL compiler.
We should unify these now that the OpenCL compiler for d3d12 has landed.