Nightly CUDA task (in AWS) fails

[aelovikov-intel]

  SYCL :: Adapters/interop-cuda-experimental.cpp
  SYCL :: Adapters/interop-experimental-single-TU-SYCL-CUDA-compilation.cpp
  SYCL :: OptionalKernelFeatures/throw-exception-for-out-of-registers-on-kernel-launch.cpp
  SYCL :: Printf/int.cpp
  SYCL :: Regression/build_log.cpp
  SYCL :: Regression/kernel_bundle_ignore_sycl_external.cpp

********************
Failed Tests (3):
  SYCL :: Basic/code_location_e2e.cpp
  SYCL :: Basic/stream/stream.cpp
  SYCL :: Graph/NativeCommand/cuda_record_async_malloc.cpp

[aelovikov-intel]

@intel/llvm-reviewers-cuda , please address both short-term (by marking the tests with unsupported/xfail/etc. as necessary) and long-term.

[EwanC]

There is a PR open with a fix for the graph fail here - #19210

[omarahmed1111]

#19232 for xfailing the other 2 tests.

[npmiller]

For reference one of the failing job: https://github.com/intel/llvm/actions/runs/15963166914/job/45021127515