Replicate member_entity_ids and policies in identity/group across nodes identically

[hghaf099]

Addresses https://hashicorp.atlassian.net/browse/VAULT-2282
Example of reading an identity/group from a standby node:

{ "request_id": "4015c4c6-ee28-eec9-a42d-a0be91970dbc", "lease_id": "", "lease_duration": 0, "renewable": false, "data": { "alias": {}, "creation_time": "2022-06-21T05:56:12.378602Z", "id": "6a769520-f4b5-59a1-d7d7-21a650c91343", "last_update_time": "2022-06-21T05:56:12.378602Z", "member_entity_ids": null, "member_group_ids": [ "66bcfa57-8aab-04c8-d9ed-263bb5d2e6b0", "b7919faf-0a16-6acd-9e4d-82ac6db946bc" ], "metadata": null, "modify_index": 1, "name": "testgroup3", "namespace_id": "root", "parent_group_ids": null, "policies": [ "default" ], "type": "internal" }, "warnings": [ "Endpoint ignored these unrecognized parameters: [-format]" ] }

A group is marshalled and persisted in the leader node and then replicated to standby nodes. In standby node, memDB is populated after unmarshalling the group entry. However, member_entity_ids is an initialized empty slice when marshalled, but become a nil entry upon unmarshalling. This creates an inconsistency on the way active/standby nodes return the results.

To fix that, it was noticed that member_entity_ids field is nil, then passed into strutil.RemoveDuplicates function so that the duplicated entries be removed. That function returns an initialized empty slice regardless of the nil argument. The fix simply checks if the member_entity_ids field is not nil before passing it to that function. A test is going to be added to the ENT repo. Further details can be found in the ticket.