[CH] Not any table column is used in the join condition.

### Backend

CH (ClickHouse)

### Bug description

[Expected behavior] and [actual behavior].

The post filter condition in join is
```
post_join_filter {
literal {
null {
bool {
nullability: NULLABILITY_NULLABLE
}
}
}
}
```
It comes from following query
```
select uid from t where uid in (select uid from t1 where partitionkey = 'xxx')
```

where `partitionkey = 'xxx'` filt out all data.

### Gluten version

_No response_

### Spark version

None

### Spark configurations

_No response_

### System information

_No response_

### Relevant logs

```bash

```

Start with the reported query and the generated post_join_filter containing a nullable null literal, then trace how the partitionkey = 'xxx' predicate is represented when it filters out all data. Done means the query no longer produces the invalid join-condition warning and its result matches the expected behavior for an empty filtered subquery.

Written by the indexing model from the issue text.