fix GS serialization issue (#4984)

Summary:
Pull Request resolved: #4984

`object_from_json` mutates its input dicts by popping `__type` keys during
deserialization. In `generation_node_from_json`, `transition_criteria` and
`generation_pausing_criteria` are decoded sequentially via `object_from_json`.

Reviewed By: saitcakmak, Balandat

Differential Revision: D95443476

fbshipit-source-id: 2d309d902ff6bff4cf5556227d39ceb7608ced21

Author: sdaulton

ax/storage/json_store/decoder.py

@@ -6,6 +6,7 @@
 
 # pyre-strict
 
+import copy
 import datetime
 import json
 from collections import OrderedDict
@@ -934,13 +935,19 @@ def generation_node_from_json(
             class_decoder_registry=class_decoder_registry,
         ),
         should_deduplicate=generation_node_json.pop("should_deduplicate", False),
+        # Deep-copy criteria JSON before decoding. object_from_json mutates
+        # its input dicts (pops "__type" keys). If the transport layer (e.g.
+        # FBLearner/msgpack) returns data with shared dict objects between
+        # transition_criteria and generation_pausing_criteria (e.g. shared
+        # TrialStatus dicts), decoding one would strip "__type" from the other,
+        # causing enums to be deserialized as plain dicts.
         transition_criteria=object_from_json(
-            object_json=transition_criteria_json,
+            object_json=copy.deepcopy(transition_criteria_json),
             decoder_registry=decoder_registry,
             class_decoder_registry=class_decoder_registry,
         ),
         pausing_criteria=object_from_json(
-            object_json=generation_pausing_criteria_json,
+            object_json=copy.deepcopy(generation_pausing_criteria_json),
             decoder_registry=decoder_registry,
             class_decoder_registry=class_decoder_registry,
         ),