Support distinct aggregation with ordering over different expressions

[findepi]

WIth DISTINCT this works

trino> SELECT array_agg(DISTINCT concat(value, value) ) csv_value
    -> FROM (VALUES 'a', 'b', 'c', 'b') t(value);
  csv_value
--------------
 [cc, aa, bb]

with ORDER BY this works

trino> SELECT array_agg( concat(value, value) ORDER BY value) csv_value
    -> FROM (VALUES 'a', 'b', 'c', 'b') t(value);
    csv_value
------------------
 [aa, bb, bb, cc]
(1 row)

with both DISTINCT and ORDER BY there is an error

trino> SELECT array_agg(DISTINCT concat(value, value) ORDER BY value) csv_value
    -> FROM (VALUES 'a', 'b', 'c', 'b') t(value);
Query 20240215_185707_00008_3tuur failed: line 1:57: For aggregate function with DISTINCT, ORDER BY expressions must appear in arguments
SELECT array_agg(DISTINCT concat(value, value) ORDER BY *value*) csv_value ...

[martint]

To elaborate on the expected semantics:

• Except for listagg, aggregation functions that allow ordering don't allow DISTINCT, and viceversa. We generalized support to apply to all aggregation functions and made it consistent with listagg behavior.
• For listagg, the rules say:

2) Let T be the argument source of AF.
...
4) Case:
  a) If <filter clause> is specified, then the <search condition>
     is effectively evaluated for each row of T. Let T1 be the collection
     of rows of T for which the result of the <search condition> is True.
  b) Otherwise, let T1 be T
...
11) If <listagg set function> is specified, then:
   a) Let T2 be T1 ordered by the <sort specification list> as specified
      in the General Rules of Subclause 10.10, “<sort specification list>”.
   b) Let TX be the single-column table that is the result of applying the 
      <character value expression> to each row of T2 and eliminating null 
      values. If one or more null values are eliminated, then a completion 
      condition is raised: warning — null value eliminated in set function (01003).
   c) Case:
      i) If DISTINCT is specified, then let TXA be the result of eliminating redundant 
         duplicate values from TX, using the comparison rules specified in Subclause 8.2, 
         “<comparison predicate>”, to identify the redundant duplicate values.
      ii) Otherwise, let TXA be TX.

In a nutshell, the argument to the aggregation is processed in this order:

• The rows in the input group are first filtered according to the FILTER (WHERE ...) clause
• The rows are then sorted based on the ORDER BY clause
• The expression in the aggregation argument is evaluated (e.g., given listagg(concat(...)), the concat(...) part is evaluated at this time)
• The results of that evaluation are deduplicated if DISTINCT is present
• The resulting values are fed into the aggregation function