make "_tidb_rowid" be able to build range

[zz-jason]

Feature Request

Is your feature request related to a problem? Please describe:

For now, _tidb_rowid can not be used to calculate the handle range, which causes a full range table scan + filter, leads to a poor performance.

TiDB(localhost:4000) > desc select * from t where _tidb_rowid = 1;
+-----------------------+----------+------+------------------------------------------------------------+
| id                    | count    | task | operator info                                              |
+-----------------------+----------+------+------------------------------------------------------------+
| Projection_4          | 8000.00  | root | test.t.a                                                   |
| └─TableReader_7       | 8000.00  | root | data:Selection_6                                           |
|   └─Selection_6       | 8000.00  | cop  | eq(test.t._tidb_rowid, 1)                                  |
|     └─TableScan_5     | 10000.00 | cop  | table:t, range:[-inf,+inf], keep order:false, stats:pseudo |
+-----------------------+----------+------+------------------------------------------------------------+
4 rows in set (0.00 sec)

Describe the feature you'd like:

make "_tidb_rowid" be able to build range, avoid the full range table scan.

[morgo]

+1 This is useful for logical dumpers that chunk split based on _tidb_rowid. I didn't even realize it was table scanning.