enable 2 test cases on XPU

[yao-matrix]

test cases

1. tests/quantization/bnb/test_mixed_int8.py::SlowBnb8bitTests::test_pipeline_cuda_placement_works_with_mixed_int8 PASS
2. tests/quantization/bnb/test_4bit.py::BnB4BitBasicTests::test_model_memory_usage FAIL, but it's ipex issue, validated success when ipex fixed, so still enable and once ipex fixed, it will naturally pass w/o modify test case.

questions

1. for tests/quantization/bnb/test_mixed_int8.py, I changed the hard-coding "cuda" to device. I tried directly it's OK to directly use torch_device, but I guess the reason you didn't use it is that maybe "rocm" system cannot work under "rocm" to(), but work on "cuda" to, so I added + device = torch_device if torch_device != "rocm" else "cuda" to workaround. This is just a guess, @hlky, let me know if you think it's not correct.

@hlky , pls help review, thx.

[DN6]

@bot /style

[github-actions]

Style fixes have been applied. View the workflow run here.

[yao-matrix]

@hlky , pls help review, thx.