[Subtask]: Implement DEDUP JOIN operator

[aunjgr]

Parent Issue

#17500

Detail of Subtask

To detect duplicate rows from data to be inserted, we will implement a DEDUP JOIN operator, which supports 3 policies on duplicate: error/ignore/update.

To help debugging, one or more new SQL keyword should also be added. The easiest way is to support queries like "select tNew.* from tNew DEDUP JOIN tOld on tNew.pk = tOld.pk".

Describe implementation you've considered

Always build on new data and probe on old data.

• on duplicate error: report error on duplicated rows
• on duplicate ignore: drop the duplicated rows
• on duplicate update: for each duplicated row update an existed row accordingly

Additional information

https://github.com/matrixorigin/docs/pull/299