[SYCL][E2E] Use system header search paths for SYCL headers in SYCL lit tests.

[tahonermann]

SYCL header files contain some uses of deprecated declarations. When SYCL headers are found via a user header search path specified by the CPATH environment variable or the -I option, these uses can cause warnings to be produced that then cause tests that use -Werror to fail. This change updates lit configuration files to use C_INCLUDE_PATH and CPLUS_INCLUDE_PATH instead of CPATH and to use -isystem instead of -I to nominate header search paths for SYCL header files.

[tahonermann]

@cperkinsintel, the Battlemage test failures don't look to me to be related to the changes I made. Both runs failed with:

ls: cannot access '/sys/kernel/debug/dri/*': No such file or directory
...
  SYCL_UR_TRACE=1 sycl-ls
...
<LOADER>[INFO]: The adapter 'libur_adapter_level_zero_v2.so.0' is skipped because UR_LOADER_USE_LEVEL_ZERO_V2 or SYCL_UR_USE_LEVEL_ZERO_V2 is not set.
<LOADER>[INFO]: failed to load adapter 'libur_adapter_cuda.so.0' with error: libcuda.so.1: cannot open shared object file: No such file or directory
<LOADER>[INFO]: failed to load adapter '/__w/llvm/llvm/toolchain/lib/libur_adapter_cuda.so.0' with error: libcuda.so.1: cannot open shared object file: No such file or directory
<LOADER>[INFO]: loaded adapter 0x0x20261d40 (libur_adapter_hip.so.0) from /__w/llvm/llvm/toolchain/lib/libur_adapter_hip.so.0
<LOADER>[INFO]: loaded adapter 0x0x2034fc90 (libur_adapter_level_zero.so.0) from /__w/llvm/llvm/toolchain/lib/libur_adapter_level_zero.so.0
<LOADER>[INFO]: loaded adapter 0x0x2034f660 (libur_adapter_native_cpu.so.0) from /__w/llvm/llvm/toolchain/lib/libur_adapter_native_cpu.so.0
<LOADER>[INFO]: loaded adapter 0x0x20352680 (libur_adapter_opencl.so.0) from /__w/llvm/llvm/toolchain/lib/libur_adapter_opencl.so.0
[native_cpu:cpu][native_cpu:0] SYCL_NATIVE_CPU, SYCL Native CPU 0.1 [0.0.0]
[opencl:cpu][opencl:0] Intel(R) OpenCL, 12th Gen Intel(R) Core(TM) i9-12900 OpenCL 3.0 (Build 0) [2025.19.3.0.17_230222]
...
[0/1] Running SYCL End-to-End tests
lit.py: /__w/llvm/llvm/llvm/sycl/test-e2e/lit.cfg.py:65: note: run-only test mode enabled, only executing tests
lit.py: /__w/llvm/llvm/llvm/sycl/test-e2e/lit.cfg.py:622: note: Targeted devices: level_zero:gpu
lit.py: /__w/llvm/llvm/llvm/sycl/test-e2e/lit.cfg.py:822: note: Found pre-installed AOT device compiler ocloc
lit.py: /__w/llvm/llvm/llvm/sycl/test-e2e/lit.cfg.py:822: note: Found pre-installed AOT device compiler opencl-aot
lit.py: /__w/llvm/llvm/llvm/sycl/test-e2e/lit.cfg.py:920: error: Cannot detect device aspect for level_zero:gpu
...
lit.py: /__w/llvm/llvm/llvm/sycl/test-e2e/lit.cfg.py:929: note: Aspects for level_zero:gpu: 
lit.py: /__w/llvm/llvm/llvm/sycl/test-e2e/lit.cfg.py:932: error: Cannot detect device SG sizes for level_zero:gpu

It looks to my untrained eye like these test runs might have gotten scheduled on a system that doesn't have the hardware needed for these tests. Can you confirm?