Ignore the legacy S3 global endpoint URL

AWS deprecated the https://s3.amazonaws.com endpoint URL we recommend
and botocore started enforcing the use of regional endpoint URLs
instead since version 1.28.0.

This updates the S3 model to ignore the legacy URL and let botocore to
calculate it from the bucket region.

storage_service/locations/models/s3.py

@@ -3,6 +3,7 @@
 import pprint
 import re
 from functools import wraps
+from urllib.parse import urlparse
 
 import boto3
 import botocore
@@ -75,10 +76,11 @@ def resource(self):
             )
             boto_args = {
                 "service_name": "s3",
-                "endpoint_url": self.endpoint_url,
                 "region_name": self.region,
                 "config": config,
             }
+            if not self._is_global_endpoint(self.endpoint_url):
+                boto_args["endpoint_url"] = self.endpoint_url
             if self.access_key_id and self.secret_access_key:
                 boto_args.update(
                     aws_access_key_id=self.access_key_id,
@@ -87,6 +89,9 @@ def resource(self):
             self._resource = boto3.resource(**boto_args)
         return self._resource
 
+    def _is_global_endpoint(self, url):
+        return urlparse(url).netloc == "s3.amazonaws.com"
+
     @boto_exception
     def _ensure_bucket_exists(self):
         """Ensure that the bucket exists by asking it something about itself.