Add Relation#unordered

[radar]

Uses Sequel's built-in unordered to remove the order clause from queries.

Today I was trying to do this query:

select { [integer::count(id).as(:count), user_id, project_id] }.group_and_count(:project_id, :user_id)

But I was told I could not do that:

[30] pry(#<TicketsRelation>)> select { [integer::count(id).as(:count), user_id, project_id] }.group_and_count(:project_id, :user_id).to_a
  ROM[postgres] (0.8ms)  SELECT "project_id", "user_id", count(*) AS "count" FROM "tickets" GROUP BY "project_id", "user_id" ORDER BY "tickets"."id"
Sequel::DatabaseError: PG::GroupingError: ERROR:  column "tickets.id" must appear in the GROUP BY clause or be used in an aggregate function
LINE 1: ...ickets" GROUP BY "project_id", "user_id" ORDER BY "tickets"....

This is because datasets are ordered by id by default.

After some googling, I found Sequel provides unordered, but rom-sql did not. Seemingly, the way to do it in rom-sql is order(nil):

select { [integer::count(id).as(:count), user_id, project_id] }.order(nil).group_and_count(:project_id, :user_id)

I think unordered is a clearer API for just a few lines of code, and it'll match what Sequel provides.

select { [integer::count(id).as(:count), user_id, project_id] }.unordered.group_and_count(:project_id, :user_id)

[solnic]

@radar thanks for the PR - could you rebase it on top of release-3.3 branch and change the base branch of this PR to release-3.3? This way I'll be able to easily release and then update master branch.

[radar]

@solnic all done