Help text for four JSON mutators promises NULL for any NULL argument, but a NULL value stores a JSON null

This is a documentation defect, not a code defect. The behaviour is
the MySQL-compatible one and is what the test suite pins; four help rows
describe it wrongly.

SELECT JSON_SET('{}', '$.a', NULL)          AS s,

       JSON_REPLACE('{"a":1}', '$.a', NULL) AS r,

       JSON_ARRAY_APPEND('[1]', '$', NULL)  AS a,

       JSON_ARRAY_INSERT('[1]', '$[0]', NULL) AS i;

s             r             a          i

{"a": null}   {"a": null}   [1, null]  [null, 1]

All four help rows say "or NULL if any of the arguments are NULL". A
NULL in the value position does not make the result NULL; it stores a
JSON null, which is the useful and expected behaviour.

How to repeat

Read the help text beside the behaviour:

SELECT JSON_SET('{"a":1}','$.b',NULL,'$.c',2) AS v;

v

{"a": 1, "b": null, "c": 2}

The wording that is right

JSON_INSERT's row (topic 368) is accurate and is the model the other
four should follow: *"returning the resulting document or NULL if either
of the json_doc or path arguments are null"*.

The four rows to correct are topics 356 (JSON_ARRAY_APPEND), 357
(JSON_ARRAY_INSERT), 386 (JSON_SET) and 389 (JSON_REPLACE), in
`scripts/fill_help_tables.sql`.

Affected versions

The four help rows have carried this wording since the functions
were documented.

Verified on 10.11 only (10.11.19-MariaDB-debug, commit 1dab253482d).
Other branches were not tested.

Note

Found while auditing the JSON functions against their documented
contracts. The code behaviour is pinned and correct -
func_json_invalid.result:360-362 records
JSON_SET('

', '$.a', NULL) ->

.