We need to extend partitioning logic and make it similar to indexing logic. We should either allow one expression (for hash partitioning), or a list of table column values and object ids (for append and range partitioning)

We need to extend partitioning logic and make it similar to indexing logic. As a first step, we should either allow one expression (for hash based partitioning), or a list of table column values and object ids (for append and range based partitioning). For both of these approaches, we can run remote min and max commands without having the need to replicate object ids to worker nodes. We can then get min and max values, and store them in a tuple or Datum.