Avoid lock if serialization result is already cached (#58722)

The GIL makes checking `self._serialize_cache is not None` atomic, so we
don't need lock.

Signed-off-by: Xinyuan <[email protected]>

Author: xinyuangui2

python/ray/air/util/tensor_extensions/arrow.py

@@ -127,6 +127,8 @@ def __init__(self, *args: Any, **kwargs: Any) -> None:
 
     def __arrow_ext_serialize__(self) -> bytes:
         """Serialize the extension type using caching if enabled."""
+        if self._serialize_cache is not None:
+            return self._serialize_cache
         with self._cache_lock:
             if self._serialize_cache is None:
                 self._serialize_cache = self._arrow_ext_serialize_compute()