`TIME_PARSE()` works for constants, fails for columns

### Affected Version

Version: 0.21.1

### Description

SQL Druid provides the [`TIME_PARSE()`](https://druid.apache.org/docs/latest/querying/sql.html#time-functions) function.

Let's try it against the `segments` table. First, verify the data, on a system with only the sample Wikipedia data:

```sql
SELECT
"start"
FROM sys.segments
LIMIT 1
```

Result:
```text
2015-09-12T00:00:00.000Z
```

Now, use `TIME_PARSE()` with the returned value. (OK, the query is kind of lame, but bear with me):

```sql
SELECT
TIME_PARSE('2015-09-12T00:00:00.000Z')
FROM sys.segments
LIMIT 1
```

Result:
```text
2015-09-12T00:00:00.000Z
```

Now, put the two together: we know the `start` column has a perfectly fine timestamp string. Try to parse it:

```sql
SELECT
TIME_PARSE("start")
FROM sys.segments
LIMIT 1
```

This fails:

```text
Error: Unknown exception
cannot translate call TIME_PARSE($t2)
java.lang.RuntimeException
```

Expected that a SQL function would work just as well for columns as for constants.

Note: I was trying to use `TIME_PARSE()` to work around [another issue](https://github.com/apache/druid/issues/11659) that prevents using `MAX()` on a string.

Reproduce the three SQL queries against the sample Wikipedia data, comparing TIME_PARSE() with the literal and the "start" column from sys.segments. Read the SQL function translation path implicated by the cannot translate call error; done means TIME_PARSE("start") returns the parsed timestamp without failure.

Written by the indexing model from the issue text.