[BUG] Aggregate Node creates new metadata for each chunk

[gva-jjoyce]

Describe the bug
Anything that processes results from a GROUP BY that goes across pages will have an inconsistent set of columns because the aggregate node creates new metadata for each page, see line 271.

This can be seen in queries which SELECT, WHERE or otherwise access columns what are the result of a GROUP BY.

To resolve, the metadata should be created one and saved to each page, similar to the mechanism used in the blob reader.

[joocer]

can replicate with new (unreleased) FAKE constructor

SELECT count(*), column_1 FROM FAKE(10000,2) GROUP BY column_1 ORDER BY COUNT(*)