`DOLT_QUERY_DIFF` falls back to keyless diff on cross-table or cross-database queries with matching schemas

When diffing two queries across different tables or databases, `DOLT_QUERY_DIFF` falls back to a keyless diff (emitting all rows as added/deleted) even when the projected columns and primary keys match.

TODO in code:
https://github.com/dolthub/dolt/blob/583165a5c83abb5be82970ce78dd56d43ed26b92/go/libraries/doltcore/sqle/dtablefunctions/dolt_query_diff.go#L214-L220

---

```sql
CREATE TABLE t (i INT PRIMARY KEY, j INT);
INSERT INTO t VALUES (1, 1), (2, 2);

CREATE TABLE tt (i INT PRIMARY KEY, j INT);
INSERT INTO tt VALUES (1, 1), (2, 3);

SELECT * FROM DOLT_QUERY_DIFF('SELECT * FROM t', 'SELECT * FROM tt');
```

```text
+--------+--------+------+------+-----------+
| from_i | from_j | to_i | to_j | diff_type |
+--------+--------+------+------+-----------+
| 1 | 1 | NULL | NULL | deleted |
| 2 | 2 | NULL | NULL | deleted |
| NULL | NULL | 1 | 1 | added |
| NULL | NULL | 2 | 3 | added |
+--------+--------+------+------+-----------+
```

Rows matching on primary key `i` should be compared and output as modified rather than treating all rows as added and deleted.

Start at go/libraries/doltcore/sqle/dtablefunctions/dolt_query_diff.go around lines 214-220, where the cross-table and cross-database fallback is marked TODO. Reproduce the behavior with the SQL example in this issue, then trace how projected columns and primary keys are handled. Done means matching primary-key rows are compared and reported as modified instead of all rows being reported as added or deleted.

Written by the indexing model from the issue text.