Enable jsonGroupArray on optional columns

[juliensagot]

We can already call jsonGroupArray() on columns sourced from a regular join, because the column is guaranteed to exist:

Reminder
  .join(RemindersList.all) { $0.remindersListID.eq($1.id) }
  .group(by: \.remindersListID)
  .select { reminder, list in
    (list.title, reminder.title.jsonGroupArray())
  }

Left joins were different: the only helper that understood optional tables worked at the table level, so if you wanted just one column you had to build the CASE … FILTER … scaffolding yourself.
This PR adds the missing column-level overload for optional tables, so the same syntax now works when the join may return NULL rows:

RemindersList
  .group(by: \.id)
  .leftJoin(Reminder.all) { $0.id.eq($1.remindersListID) }
  .select { list, reminder in
    (list.title, reminder.title.jsonGroupArray())
  }

Tests cover the new overload alongside the existing distinct/order/filter behavior.

[stephencelis]

@juliensagot Did you mean to close this without follow-up?

[juliensagot]

@stephencelis Yes! That syntax is already supported by the current version of the library, so that PR is useless 😅