SQL `max()`, `min()` fail for `VARCHAR` (`string`) columns

### Affected Version

Version: 0.21.1

### Description

The SQL functions `min()` and `max()` are typically defined to operate over any type with an ordering. However, Druid fails when they are used with a `VARCHAR` (Druid `string`) column. Example:

```sql
SELECT
min("start"), max("end")
FROM sys.segments
```

Results:

```text
Error: Unknown exception
java.lang.reflect.InvocationTargetException
java.lang.RuntimeException
```

But, the following does work:

```sql
SELECT
min("size"), max("size")
FROM sys.segments
```

When run against the sample Wikipedia data, we get a more detailed error message:

```sql
SELECT min("user")
FROM "wikipedia"
```

```text
Error: Unknown exception

Error while applying rule DruidQueryRule(AGGREGATE), args [rel#1235:LogicalAggregate.NONE.[]
(input=RelSubset#1230,group={},EXPR$0=MIN($19)), rel#1244:DruidQueryRel.NONE.[](query=
{"queryType":"scan","dataSource":{"type":"table","name":"wikiticker-2015-09-12-sampled"},"intervals":
{"type":"intervals","intervals":["2021-09-03T23:04:08.000Z/146140482-04-24T15:36:27.903Z"]},"virtualColumns":
[],"resultFormat":"compactedList","batchSize":20480,"order":"none","filter":null,"columns":
["__time","added","channel","cityName","comment","countryIsoCode","countryName","deleted","delta","
isAnonymous","isMinor","isNew","isRobot","isUnpatrolled","metroCode","namespace","page","
regionIsoCode","regionName","user"],"legacy":false,"context":{"sqlOuterLimit":100,
"sqlQueryId":"531a8bef-ce4b-45a7-96f4-c153d98a0951"},"descending":false,"
granularity":{"type":"all"}},signature={__time:LONG, added:LONG, channel:STRING,
cityName:STRING, comment:STRING, countryIsoCode:STRING, countryName:STRING,
deleted:LONG, delta:LONG, isAnonymous:STRING, isMinor:STRING, isNew:STRING,
isRobot:STRING, isUnpatrolled:STRING, metroCode:STRING, namespace:STRING,
page:STRING, regionIsoCode:STRING, regionName:STRING, user:STRING})]

java.lang.RuntimeException
```

### Workaround

Tried converting the `start` value to a `TIMESTAMP` with `TIME_PARSE()`, but that [failed](https://github.com/apache/druid/issues/11660).

We can, however, issue a separate query for each value, and use `ORDER BY` to do the ordering. For the minimum:

```sql
SELECT "start"
FROM sys.segments
ORDER BY "start" ASC
LIMIT 1
```

This approach does not scale if we need to work with many columns, or if we need the answer from a single query.

No source file or test is named. Start by reproducing the min() and max() queries against sys.segments and the sample Wikipedia data, comparing VARCHAR and numeric columns; done means both functions work for string columns without regressing numeric aggregation.

Written by the indexing model from the issue text.