merge NPU(Ascend) backend logic in Int4PlainInt32Tensor subclass

Author: orangeH25

test/quantization/quantize_/workflows/int4/test_int4_plain_int32_tensor.py

@@ -100,8 +100,8 @@ def test_activation_prescaling(self):
 
 @unittest.skipIf(not torch_version_at_least("2.7.1"), "Need pytorch 2.7.1+")
 @unittest.skipIf(
-    torch.accelerator.current_accelerator().type == "npu"
-    and torch.accelerator.is_available(),
+    torch.accelerator.current_accelerator().type != "npu"
+    or not torch.accelerator.is_available(),
     "NPU not available",
 )
 class Int4PlainInt32TensorNPU(TestCase):