[SYCL][E2E] FreeFunctionKernels/enqueue_functions_sub_group_size.cpp fails on CI

### Describe the bug

FreeFunctionKernels/enqueue_functions_sub_group_size.cpp fails on CI
see example output https://github.com/intel/llvm/actions/runs/30254701498/job/89946582190?pr=22771:

```

FAIL: SYCL :: FreeFunctionKernels/enqueue_functions_sub_group_size.cpp (1500 of 2564)
******************** TEST 'SYCL :: FreeFunctionKernels/enqueue_functions_sub_group_size.cpp' FAILED ********************
Exit Code: -6

Command Output (stdout):
--
# RUN: at line 5
/__w/llvm/llvm/toolchain/bin/dpclang++ -Werror -fsycl -fsycl-targets=native_cpu /__w/llvm/llvm/llvm/sycl/test-e2e/FreeFunctionKernels/enqueue_functions_sub_group_size.cpp -o /__w/llvm/llvm/build-e2e/FreeFunctionKernels/Output/enqueue_functions_sub_group_size.cpp.tmp.out
# executed command: /__w/llvm/llvm/toolchain/bin/dpclang++ -Werror -fsycl -fsycl-targets=native_cpu /__w/llvm/llvm/llvm/sycl/test-e2e/FreeFunctionKernels/enqueue_functions_sub_group_size.cpp -o /__w/llvm/llvm/build-e2e/FreeFunctionKernels/Output/enqueue_functions_sub_group_size.cpp.tmp.out
# note: command had no output on stdout or stderr
# RUN: at line 6
env ONEAPI_DEVICE_SELECTOR=native_cpu:cpu /__w/llvm/llvm/build-e2e/FreeFunctionKernels/Output/enqueue_functions_sub_group_size.cpp.tmp.out
# executed command: env ONEAPI_DEVICE_SELECTOR=native_cpu:cpu /__w/llvm/llvm/build-e2e/FreeFunctionKernels/Output/enqueue_functions_sub_group_size.cpp.tmp.out
# .---command stderr------------
# | terminate called after throwing an instance of 'sycl::_V1::exception'
# | what(): Sub-group size 16 is not supported on the device
# `-----------------------------
# error: command failed with exit status: -6
```

### To reproduce

### Environment

- OS: Linux
- Target device and vendor: Intel GPU

### Additional context

_No response_

The failing test is llvm/sycl/test-e2e/FreeFunctionKernels/enqueue_functions_sub_group_size.cpp. Reproduce it with the dpclang++ command and ONEAPI_DEVICE_SELECTOR=native_cpu:cpu invocation shown in the CI output, then inspect how the test requests sub-group size 16 and how native_cpu reports device support. Done means the failure is resolved and this E2E test passes on the affected CI target.

Written by the indexing model from the issue text.