Fix the null handling for `to_char` function

[goldmedal]

Describe the bug

Currenlty, if we input a null value to to_char, we will get an empty string instead of a null value.

> select to_char(NULL, '%Y-%m-%d %H:%M:%S') is null;
+-------------------------------------------------+
| to_char(NULL,Utf8("%Y-%m-%d %H:%M:%S")) IS NULL |
+-------------------------------------------------+
| false                                           |
+-------------------------------------------------+
1 row(s) fetched. 
Elapsed 0.001 seconds.

> select to_char(NULL, '%Y-%m-%d %H:%M:%S');
+-----------------------------------------+
| to_char(NULL,Utf8("%Y-%m-%d %H:%M:%S")) |
+-----------------------------------------+
|                                         |
+-----------------------------------------+
1 row(s) fetched. 
Elapsed 0.001 seconds.

However, the postgres behavior is returing a null value if got a null input.

psql=# select to_char(null::timestamp, '%Y-%m-%d %H:%M:%S') is null;
 ?column? 
----------
 t
(1 row)

psql=# select to_char(null::timestamp, '%Y-%m-%d %H:%M:%S');
 to_char 
---------
 
(1 row)

I think we should follow the behavior of postgres.

To Reproduce

Execute the SQL mentioned above.

Expected behavior

to_char should return a null if got a null input.

Additional context

No response

[kosiew]

follow the behavior of postgres.

Per the current test, this is working by design:

datafusion/datafusion/sqllogictest/test_files/timestamps.slt

Lines 2846 to 2848 in 3193014

	SELECT to_char(null, '%d-%m-%Y');
	----
	(empty)

[goldmedal]

follow the behavior of postgres.

Per the current test, this is working by design:

Yes, but it doesn't make sense to me. I prefer to follow how Postgres handles this.

[Omega359]

I went looking for a system where null input would return an empty string and came up blank. I am fairly certain spark would return null, and the equiv function in duckdb returns null

D  select strftime(null::timestamp, '%Y-%m-%d %H:%M:%S');
┌────────────────────────────────────────────────────────┐
│ strftime(CAST(NULL AS TIMESTAMP), '%Y-%m-%d %H:%M:%S') │
│                        varchar                         │
├────────────────────────────────────────────────────────┤
│ NULL                                                   │
└────────────────────────────────────────────────────────┘