[3/N] Synchronize test files with upstream generalization changes

[pbielak]

The recent upstream generalization changes need to be synchronized with torch-xpu-ops. This PR considers files that were not present in torch-xpu-ops - add entries to skip_list_common.py to run them:

• test_ao_sparsity.py ([3/N][Test] Enable torch_ao_sparsity.py for XPU pytorch/pytorch#191202)
• autograd/test_functional.py ([Test] Generalize autograd/test_functional.py tests and enable on XPU pytorch/pytorch#191342)
• test_functional_autograd_benchmark.py ([Test] Generalize test_functional_autograd_benchmark.py pytorch/pytorch#191474)
• test_mkldnn.py ([Test] Generalize test_mkldnn.py tests and enable on XPU pytorch/pytorch#191740)
• lazy/test_reuse_ir.py ([Test][XPU] Refactor reuse_ir tests to device-type harness and enable XPU pytorch/pytorch#192240)
• test_stateless.py ([Test] Move device-dependent tests in test_stateless.py to a device-generic class pytorch/pytorch#192218)
• test_overrides.py ([Test] Add hw_classification and enable XPU for test_overrides.py pytorch/pytorch#194257)
• functorch/test_compile_to_python.py ([Test] Make test/functorch/test_compile_to_python.py tests device-generic and enable XPU. pytorch/pytorch#195457)
• test_opaque_obj_v2.py ([Test] Refactor test_opaque_obj_v2.py and enable tests on XPU pytorch/pytorch#195854)

[BBBela]

It makes sense for me to include those new test files in torch-xpu-ops CI, especially as they are now generalized in upstream and launched in upstream XPU CI.
It looks like necessary step to prevent all the torch-xpu-ops PRs from introducing any regression.

Although they were not ported for XPU so far due to reasons I don't know, so it would be best to have an official confirmation from PRC team that we can proceed with enabling testing all the device-related test files from upstream that are already generic, but have not been tested in torch-xpu-ops CI so far.

To sum up - for me the change looks good, but I will wait with approval until PRC team confirms we can introduce all those new tests into torch-xpu-ops CI.