Fixes when using aggregates inside Spread Embedded Resources

[laurenceisla]

Missing:

• Tests
• Refactor
• Changelog

These fixes are also needed for #3640

[laurenceisla]

For Spread Resources I decided to use COUNT(<row>) instead of COUNT(*). When there are nested spread resources, doing a COUNT(*) may not return the expected values, since it would count all the columns * from the intermediate resources.

For example, for this query:

curl 'localhost:3000/process_supervisor?select=...processes(factory_id,...process_costs(count()))'

The resulting SQL would be:

SELECT "process_supervisor_processes_1"."factory_id",
       COUNT("process_supervisor_processes_1"."count") AS "count"
FROM "test"."process_supervisor"
LEFT JOIN LATERAL
  (SELECT "processes_1"."factory_id",
          "processes_process_costs_2"."count"
   FROM "test"."processes" AS "processes_1"
   LEFT JOIN LATERAL
     (SELECT "process_costs_2" AS "count"
      FROM "test"."process_costs" AS "process_costs_2"
      WHERE "process_costs_2"."process_id" = "processes_1"."id") AS "processes_process_costs_2" ON TRUE
   WHERE "processes_1"."id" = "test"."process_supervisor"."process_id") AS "process_supervisor_processes_1" ON TRUE
GROUP BY "process_supervisor_processes_1"."factory_id"

Using COUNT("process_supervisor_processes_1".*) would not count correctly since there are processes that do not have process_count assigned.

[steve-chavez]

LGTM!

[laurenceisla]

Did not backport the fixes to v12 because the last one depends on a feature/performance commit, so it has a conflict with the code there.