Invalid characters in MC break search API

Nobody has claimed this yet.

Adding messages like this to MC seems to break the ability of the client to fetch them:
'msg': '[\x02L\u200bi\u200bl\u200bB\u200br\u200bu\u200bc\u200be\x02] sss'

Indexing with the client works fine:
{'error': None,
'items': [{'bulk': {'_id': 1370872170497990695,
'_index': 'main',
'created': 1,
'deleted': 0,
'result': 'created',
'status': 201,
'updated': 0}}]}

However, on search, these are the results:
{'took': None, 'timed_out': None, 'aggregations': None, 'hits': None, 'profile': None, 'warning': None}

In the server query log at the same time, I can see that this did in fact return results the client is not displaying.
/* Thu Sep 8 14:54:48.419 2022 conn 2 real 0.001 wall 0.002 found 73

Using requests seems to work:
response = requests.post(f"{settings.MANTICORE_URL}/json/search", json=search_query)
response = ujson.loads(response.text)

Start by reproducing the search request with the message containing control and zero-width characters, comparing the Python client's response with the working requests and ujson path shown in the issue. Trace how the client parses the search response and verify that the returned hits are exposed instead of becoming None.

Written by the indexing model from the issue text.