[Test][XPU] Added gpu cache cleaning for XPU devices (#917)

## Summary
Enable GPU cache cleaning for XPU devices


## Details
I discovered this issue while working on
intel/intel-xpu-backend-for-triton#5292
Basically testing on XPU was hanging indefinitely, but it was not due to
a specific test. Hanging only occurred if many tests were run
sequentially. In my local experiments on Intel Data Center GPU Max 1100
this patch resolves hanging issue.

The patch is small and unlikely to cause problems for cuda.

## Testing Done
`pytest test` with Intel Data Center GPU Max 1100

- Hardware Type: Intel Data Center GPU Max 1100
- [x] run `make test` to ensure correctness
- [x] run `make checkstyle` to ensure code style
- [x] run `make test-convergence` to ensure convergence

Author: Egor-Krivov

test/conftest.py

@@ -3,6 +3,9 @@
 
 
 @pytest.fixture(autouse=True)
-def clear_cuda_cache():
+def clear_gpu_cache():
     yield
-    torch.cuda.empty_cache()
+    if torch.cuda.is_available():
+        torch.cuda.empty_cache()
+    elif torch.xpu.is_available():
+        torch.xpu.empty_cache()