arrivals_departures/connections: move id column into separate view

[derhuerst]

Currently, queries on arrivals_departures & connections are slow as soon as there are frequencies-based rows in there. This is because the frequencies_it – which is needed for arrival_departure_id/connection_id – calculation uses a row_number() over a PARTITION BY.

This PR

• moves these expensive columns into separate views arrivals_departures_with_ids/connections_with_ids;
• adapts arrival_departure_by_arrival_departure_id/connection_by_connection_id to use the new views;
• adds benchmarking for the case described above.

[derhuerst]

Currently, this still causes different sets of arrivals/departures for the (functionally) same query.

git checkout main
# import data...
psql -c 'SELECT route_id, route_short_name, trip_id, trip_headsign, frequencies_it, stop_sequence_consec, t_arrival, t_departure, stop_id, stop_name, station_id FROM arrivals_departures WHERE trip_id = \'171756466\' AND date = \'2022-08-23\' AND t_departure >= \'2022-08-23T06:10+01:00\' AND t_departure <= \'2022-08-23T06:24+01:00\' ORDER BY route_id, trip_id, date, frequencies_it, stop_sequence_consec' | sponge before
git checkout arrs-deps-cons-ids
# import data...
psql -c 'SELECT route_id, route_short_name, trip_id, trip_headsign, frequencies_it, stop_sequence_consec, t_arrival, t_departure, stop_id, stop_name, station_id FROM arrivals_departures_with_ids WHERE trip_id = \'171756466\' AND date = \'2022-08-23\' AND t_departure >= \'2022-08-23T06:10+01:00\' AND t_departure <= \'2022-08-23T06:24+01:00\' ORDER BY route_id, trip_id, date, frequencies_it, stop_sequence_consec' | sponge after
diff before after