[SPARK-40308][SQL] Allow non-foldable delimiter arguments to str_to_map function

[bersprockets]

What changes were proposed in this pull request?

Remove the check for foldable delimiter arguments from StringToMap#checkInputDataTypes.

Except for checkInputDataTypes, StringToMap is already able to handle non-foldable delimiter arguments (no other changes are required).

Why are the changes needed?

Hive 2.3.9 allows non-foldable delimiter arguments, e.g.:

drop table if exists maptbl;
create table maptbl as select ',' as del1, ':' as del2, 'a:1,b:2,c:3' as str;
insert into table maptbl select '%' as del1, '-' as del2, 'a-1%b-2%c-3' as str;
select str, str_to_map(str, del1, del2) from maptbl;

This returns

+--------------+----------------------------+
|     str      |            _c1             |
+--------------+----------------------------+
| a:1,b:2,c:3  | {"a":"1","b":"2","c":"3"}  |
| a-1%b-2%c-3  | {"a":"1","b":"2","c":"3"}  |
+--------------+----------------------------+
2 rows selected (0.13 seconds)

However, Spark returns an error:

str_to_map's delimiters must be foldable.; line 1 pos 12;

The use-case is more likely to be something like this:

select
  str,
  str_to_map(str, ',', if(region = 0, ':', '#')) as m
from
  maptbl2;

Does this PR introduce any user-facing change?

Yes, users can now specify non-foldable delimiter arguments to str_to_map.

Literals are still accepted, so the change is backwardly compatible.

How was this patch tested?

New unit tests.

[bersprockets]

Not really a test of my change, but it was a missing test (the case where a literal pair delimiter is given but no key/value delimiter is given).

[MaxGekk]

@cloud-fan I wonder why did you insist in #13990 (comment) that the delimiters must be foldable.

[MaxGekk]

If you invoke only the parent method, let's remove checkInputDataTypes() then.

[MaxGekk]

+1, LGTM. Merging to master.
Thank you, @bersprockets and @HyukjinKwon for review.