Mark ARRAY_AGG(DISTINCT ...) not implemented

[james727]

Which issue does this PR close?

This is the last change required to close #1512

Rationale for this change

Right now array_agg(distinct ...) doesn't work. The physical plan construction logic uses the non-distinct array_agg whether or not distinct was specified. Interestingly enough it still works correctly under certain conditions, due to the SingleDistinctToGroupBy optimizer rule.

As an example, consider the following queries:

--- Works, since logical plan is rewritten with a subquery and non-distinct agg. Will return:
--- +------------------------------------------+
--- | ARRAYAGG(DISTINCT aggregate_test_100.c2) |
--- +------------------------------------------+
--- | [2, 3, 5, 1, 4]                          |
--- +------------------------------------------+
SELECT array_agg(DISTINCT c2) FROM aggregate_test_100;

--- Returns incorrect results, since SingleDistinctToGroupBy optimizer rule does not apply:
--- +--------------------------------------------------------------------------+
--- | ARRAYAGG(DISTINCT aggregate_test_100.c2)      | COUNT(DISTINCT UInt8(1)) |
--- +--------------------------------------------------------------------------+
--- | [2, 5, 1, 1, 5, 4, 3, 3, 1, 4, 1, 4, 3, ...]  | 1                        |
--- +--------------------------------------------------------------------------+
SELECT array_agg(DISTINCT c2), count(distinct 1) FROM aggregate_test_100;

After this change distinct array agg will throw an error when SingleDistinctToGroupBy does not apply.

I'm planning on working on actually implementing distinct array_agg after this, but figured this was worth fixing for now.

What changes are included in this PR?

This marks the aggregation not implemented, and adds a block for testing count/approx distinct/array agg with distinct = true.

[alamb]

Looks good to me -- thank you @james727

do distinct aggregates make sense for array aggregates? If they make sense could you possibly file at ticket to track implementing them?

[james727]

@alamb yep - they do! I think #1323 covers it but I can add another ticket and close the set_agg one if easier.