Use memoryviews for writes only for Python 3.10+

The core SDK validates that if a bytes iterable is provided that
the object has an __iter__() method available:
https://github.com/Azure/azure-sdk-for-python/blob/08e0cd870062e11e22998d6e10ef55581b35c045/sdk/core/azure-core/azure/core/rest/_helpers.py#L146

However memoryview objects did not expose an __iter__() until Python
3.10 via: python/cpython#22119

We still want to leverage memoryviews when possible to reduce the
number of potential copies so we wrap the bytearrays sent for
stage_blocks if we are on Python 3.10+. Otherwise, we just directly
provide the bytearray as is.

Author: kyleknap

src/azstoragetorch/_client.py

@@ -11,6 +11,7 @@
 import math
 import os
 import random
+import sys
 import threading
 import time
 import urllib.parse
@@ -197,6 +198,11 @@ def stage_blocks(
     ) -> List[STAGE_BLOCK_FUTURE_TYPE]:
         if not data:
             raise ValueError("Data must not be empty.")
+        if (
+            isinstance(data, memoryview)
+            and not self._sdk_supports_memoryview_for_writes()
+        ):
+            data = data.obj
         stage_block_partitions = self._get_stage_block_partitions(data)
         futures = []
         for pos, length in stage_block_partitions:
@@ -371,3 +377,12 @@ def _get_url_without_query_string(
                 None,
             )
         )
+
+    def _sdk_supports_memoryview_for_writes(self) -> bool:
+        # The SDK validates objects passed to its HTTP request layer expose an __iter__()
+        # method that can be used to iterate through bytes passed to it. However, memoryview
+        # objects did not expose an __iter__() method till Python 3.10.
+        #
+        # We still want to leverage memorviews when we can to avoid unnecessary copies. So
+        # we check the Python version to determine if we can use memoryviews for writes.
+        return sys.version_info >= (3, 10)