Introduce ordered set aggregate functions

[beikov]

The idea is to extend HQL to accept the syntax function_name(DISTINCT? args*) WITHIN GROUP (ORDER BY ...) for functions that are ordered set aggregate functions. Here is a nice explanation of ordered set aggregate functions and specifically, the interesting listagg variant for which we need a special grammar rule because it also has an ON OVERFLOW clause: https://modern-sql.com/feature/listagg

The emulation of the listagg function can easily be done on most databases through the use of the various vendor specific versions e.g. string_agg and group_concat. Emulating the function with a recursive CTE is also possible but probably not practical, so for databases that do not support this function, we should simply provide an error.

Out of the databases that we have in hibernate-core (the others are now in hibernate-community-dialects), the following databases do not support the string aggregation:

• Apache Derby
• Sybase ASE

It might be possible to implement UDFs to emulate this function, but that is out of scope.

Apart from the standard listagg function, I propose we also add the following functions (without emulation, at least for now) which are nicely explained in the PostgreSQL documentation (https://www.postgresql.org/docs/13/functions-aggregate.html#FUNCTIONS-ORDEREDSET-TABLE):

• mode
• percentile_cont
• percentile_disc

And the following hypothetical set aggregate functions (https://www.postgresql.org/docs/13/functions-aggregate.html#FUNCTIONS-HYPOTHETICAL-TABLE):

• rank
• dense_rank
• percent_rank
• cume_dist

[beikov]

See #4758