[CI] Patch torch.library.infer_schema for fused moe ops to fix CI (#854)

make sure pytorch infer_schema check is patched before some case which
using fused moe ops:
1. model register
2. quantization loading
3. fused moe ut

Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>

tests/ops/test_fused_moe.py

@@ -19,6 +19,9 @@
 
 Run `pytest tests/ops/test_fused_moe.py`.
 """
+# fused moe ops test will hit the infer_schema error, we need add the patch
+# here to make the test pass.
+import vllm_ascend.patch.worker.patch_common.patch_utils  # type: ignore[import]  # isort: skip  # noqa
 
 import pytest
 import torch