add missing Mma-specialization for m16n8k32.s32.s4.s4.s32

[dxqb]

there appear to be missing specializations for certain operator shapes. This leads to an incomplete type:

using ElementA = cutlass::int4b_t;
using ElementB = cutlass::int4b_t;
using ElementC = int32_t;
using ElementAccumulator = int32_t;
using LayoutA = cutlass::layout::RowMajor;
using LayoutB = cutlass::layout::ColumnMajor;
using LayoutC = cutlass::layout::RowMajor;
using MMAOp  = cutlass::arch::OpClassTensorOp;
using SmArch = cutlass::arch::Sm80;
using OperatorShape    = cutlass::gemm::GemmShape<16, 8, 32>;
...
using Gemm = cutlass::gemm::device::Gemm<...

But there is m16n8k32 instruction for s4.

The attached code seems to work, but take this with a grain of salt because I'm very new to CUTLASS. It's also incomplete - if this is correct, it also needs to be added for u4 and its combinations.