[XPU] Changed how XPU discovery works during setup.py (#720)

## Summary
Right now we check `xpu-smi` during installation to find out if machine
has XPU device. But `xpu-smi` is often missing from user devices, so we
end up discovering incorrect platform (`cpu`) and then try to install
wrong `triton` dependency. User ends up with `triton-xpu` being
overwritten by `triton`. This combination doesn't work.

So I check `sycl-ls` which should be available. The output I get on PVC
machine:
```
[level_zero:gpu][level_zero:0] Intel(R) oneAPI Unified Runtime over Level-Zero, Intel(R) Data Center GPU Max 1100 12.60.7 [1.6.32567+18]
[opencl:cpu][opencl:0] Intel(R) OpenCL, Intel(R) Xeon(R) Gold 6438Y+ OpenCL 3.0 (Build 0) [2024.18.12.0.05_160000]
[opencl:gpu][opencl:1] Intel(R) OpenCL Graphics, Intel(R) Data Center GPU Max 1100 OpenCL 3.0 NEO  [25.05.32567]
```

The output I get on B570 machine:
```
[level_zero:gpu][level_zero:0] Intel(R) oneAPI Unified Runtime over Level-Zero, Intel(R) Arc(TM) B570 Graphics 20.1.0 [1.6.32567+19]
[opencl:cpu][opencl:0] Intel(R) OpenCL, Intel(R) Core(TM) Ultra 7 265K OpenCL 3.0 (Build 0) [2025.19.4.0.18_160000.xmain-hotfix]
[opencl:gpu][opencl:1] Intel(R) OpenCL Graphics, Intel(R) Arc(TM) B570 Graphics OpenCL 3.0 NEO  [25.05.32567]
```

## Possible alternative

We could just import pytorch and check `torch.xpu.is_available()`. It
might even be better that way, since right now if user is missing torch
we will try to install cuda torch even for XPU devices.

## Testing Done
I tested it by installing on new conda environment from the source code
with:
```
pip install torch==2.7.0 torchvision==0.22.0 torchaudio==2.7.0 --index-url https://download.pytorch.org/whl/xpu --no-cache-dir
pip install -e .
```

- Hardware Type: XPU
- [ ] run `make test` to ensure correctness
- [ ] run `make checkstyle` to ensure code style
- [ ] run `make test-convergence` to ensure convergence

Author: Egor-Krivov

setup.py

@@ -47,6 +47,27 @@ def get_optional_dependencies():
     }
 
 
+def is_xpu_available():
+    """
+    Check if Intel XPU is available.
+    xpu-smi is often missing right now.
+    """
+    try:
+        subprocess.run(["xpu-smi"], check=True)
+        return True
+    except (subprocess.SubprocessError, FileNotFoundError):
+        pass
+
+    try:
+        result = subprocess.run("sycl-ls", check=True, capture_output=True, shell=True)
+        if 'level_zero:gpu' in result.stdout.decode():
+            return True
+    except (subprocess.SubprocessError, FileNotFoundError):
+        pass
+
+    return False
+
+
 def get_platform() -> Literal["cuda", "rocm", "cpu", "xpu"]:
     """
     Detect whether the system has NVIDIA or AMD GPU without torch dependency.
@@ -63,11 +84,10 @@ def get_platform() -> Literal["cuda", "rocm", "cpu", "xpu"]:
             print("ROCm GPU detected")
             return "rocm"
         except (subprocess.SubprocessError, FileNotFoundError):
-            try:
-                subprocess.run(["xpu-smi"], check=True)
+            if is_xpu_available():
                 print("Intel GPU detected")
                 return "xpu"
-            except (subprocess.SubprocessError, FileNotFoundError):
+            else:
                 print("No GPU detected")
                 return "cpu"