add int8/tf32 transpose A copy traits

[taozha2]

No description provided.

[taozha2]

@aacostadiaz @joeatodd This PR add int8 and tf32 transpose copy traits support including both A and B. all transpose case for bf16/fp16/int8/uint8/tf32 here https://github.com/taozha2/cutlass-fork/blob/zt/gemm_layout_data_type/examples/sycl/pvc/pvc_gemm.cpp#L432 passed(MUST based on latest driver https://ubit-gfx.intel.com/build/21406574 or later)..
But as i have talked with @aacostadiaz before, latest code base have regression for int8/tf32 transpose gemm support, i think all transpose case should be passed if you fixed that regression.
please review this PR and merge it.

[taozha2]

All below case should be pass with the new intrinsic in this PR after the regression fixed.
https://github.com/codeplaysoftware/cutlass-fork/blob/sycl-develop/test/unit/gemm/device/xe_gemm_s8_s8_s32_tensor_op_s32.cpp#L71
https://github.com/codeplaysoftware/cutlass-fork/blob/sycl-develop/test/unit/gemm/device/xe_gemm_s8_s8_s32_tensor_op_s32_cooperative.cpp#L83
https://github.com/codeplaysoftware/cutlass-fork/blob/sycl-develop/test/unit/gemm/device/xe_gemm_tf32_tf32_fp32_tensor_op_fp32.cpp#L45
https://github.com/codeplaysoftware/cutlass-fork/blob/sycl-develop/test/unit/gemm/device/xe_gemm_tf32_tf32_fp32_tensor_op_fp32_cooperative.cpp#L85

this is the passed case before regression:
fb7df75#diff-710a282c2eba335641e0a4017b9167e937676d901ab525b010b23cc309e04a8aR155
fb7df75#diff-710a282c2eba335641e0a4017b9167e937676d901ab525b010b23cc309e04a8aR193

[taozha2]

@aacostadiaz @mehdi-goli can you merge this PR?

[t4c1]

@taozha2 can you merge in sycl-develop branch?

[taozha2]

@taozha2 can you merge in sycl-develop branch?

@mehdi-goli @aacostadiaz @t4c1 I have rebased the PR, you can review it

[joeatodd]

I would prefer to avoid the _cacheopts_ variant since we don't use it elsewhere, but if it's needed for some reason then 👍