`02-fused-softmax` tutorial fails on BMG

[anmyachev]

CI: https://github.com/intel/intel-xpu-backend-for-triton/actions/runs/15361783058/job/43231263982

****** Running 02-fused-softmax test ******

+ run_tutorial_args+=("--reports=$TRITON_TEST_REPORTS_DIR")
+ python -u /runner/_work/intel-xpu-backend-for-triton/intel-xpu-backend-for-triton/scripts/run_tutorial.py --skip-list=/runner/_work/intel-xpu-backend-for-triton/intel-xpu-backend-for-triton/scripts/skiplist/xe2/tutorials.txt 02-fused-softmax.py --reports=/runner/_work/intel-xpu-backend-for-triton/intel-xpu-backend-for-triton/reports
Traceback (most recent call last):
  File "/runner/_work/intel-xpu-backend-for-triton/intel-xpu-backend-for-triton/scripts/run_tutorial.py", line 133, in <module>
    main()
  File "/runner/_work/intel-xpu-backend-for-triton/intel-xpu-backend-for-triton/scripts/run_tutorial.py", line 125, in main
    time = run_tutorial(tutorial_path)
  File "/runner/_work/intel-xpu-backend-for-triton/intel-xpu-backend-for-triton/scripts/run_tutorial.py", line 67, in run_tutorial
    spec.loader.exec_module(module)
  File "<frozen importlib._bootstrap_external>", line 850, in exec_module
  File "<frozen importlib._bootstrap>", line 228, in _call_with_frames_removed
  File "02-fused-softmax.py", line 234, in <module>
    benchmark.run(show_plots=True, print_data=True)
  File "/runner/_work/intel-xpu-backend-for-triton/intel-xpu-backend-for-triton/scripts/run_tutorial.py", line 30, in run
    result = super().run(save_path=tmp_dir, **kwargs)
  File "/opt/hostedtoolcache/Python/3.9.22/x64/lib/python3.9/site-packages/triton/testing.py", line 470, in run
    result_dfs.append(self._run(bench, save_path, show_plots, print_data, **kwargs))
  File "/opt/hostedtoolcache/Python/3.9.22/x64/lib/python3.9/site-packages/triton/testing.py", line 417, in _run
    ret = self.fn(**x_args, **{bench.line_arg: y}, **bench.args, **kwrags)
  File "02-fused-softmax.py", line 229, in benchmark
    ms = triton.testing.do_bench(lambda: softmax(x))
  File "/opt/hostedtoolcache/Python/3.9.22/x64/lib/python3.9/site-packages/triton/testing.py", line 225, in do_bench
    end_event.record()
  File "/opt/hostedtoolcache/Python/3.9.22/x64/lib/python3.9/site-packages/torch/xpu/streams.py", line 129, in record
    super().record(stream)
RuntimeError: Native API failed. Native API returns: 39 (UR_RESULT_ERROR_OUT_OF_DEVICE_MEMORY)