microsoft/compiler: Hook up shader model 6.8 with new comparison sampling intrinsics
Shader model 6.8 is in the D3D headers and in DXC, so go ahead and hook it up in our compiler stack and drivers. The primary relevant bit for 6.8 that we couldn't emulate in a slower way is the new comparison sampling intrinsics from https://github.com/microsoft/hlsl-specs/pull/147. So go ahead and hook those up.
Once we have a D3D+WARP+DXIL.dll in CI that can use SM6.8, Dozen's fail list should go down to 0 (on the snapshot CTS that we're currently using).