GROUP BY constant with aggregation function and with empty input magically summons a row

### Describe the bug

Hello, it seems that there is inconsistent behaviour for select queries with group by on constant when there are no input rows.

When I have no aggregation functions then result of such query is empty, but as soon as I add some aggregation function I get single row result.

### To Reproduce

Run following query in datafusion-cli
```
> select 'wat' as key, count(1) as count from (select * from unnest(array[]::text[])) as b(field) group by key;
+-----+-------+
| key | count |
+-----+-------+
| wat | 0 |
+-----+-------+
1 row(s) fetched.
Elapsed 0.004 seconds.
```

### Expected behavior

Empty result

### Additional context

Query with no aggregation functions gives empty result
```
DataFusion CLI v46.0.1
> select 'wat' as key from (select * from unnest(array[]::text[])) as b(field) group by key;
+-----+
| key |
+-----+
+-----+
0 row(s) fetched.
Elapsed 0.004 seconds.
```

Postgres gives empty results in both cases
```
psql (17.4 (Homebrew))
Type "help" for help.

postgres=# select 'wat' as key from (select * from unnest(array[]::text[])) as b(field) group by key;
key
-----
(0 rows)

postgres=# select 'wat' as key, count(1) as count from (select * from unnest(array[]::text[])) as b(field) group by key;
key | count
-----+-------
(0 rows)
```

Start by reproducing the two SQL queries in datafusion-cli v46.0.1 and compare their results with PostgreSQL. Trace how empty input, a constant GROUP BY key, and COUNT aggregation are handled; done means the aggregation query returns zero rows for this case, with regression coverage for the behavior.

Written by the indexing model from the issue text.