fix(orchestrator): slightly improve the timeout query

[TBonnin]

simplifying slightly the where clause selecting tasks that have timeout out from

CREATED AND created_timeout < NOW
OR STARTED AND heartbeat_timeout < NOW
OR STARTED AND completed_timeout < NOW

to

CREATED AND created_timeout < NOW
OR (STARTED AND (heartbeat_timeout < NOW OR completed_timeout < NOW`))

Query plan before

Update on tasks  (cost=17.97..26.04 rows=1 width=91)
  ->  Nested Loop  (cost=17.97..26.04 rows=1 width=91)
        ->  HashAggregate  (cost=17.41..17.42 rows=1 width=56)
              Group Key: "ANY_subquery".id
              ->  Subquery Scan on "ANY_subquery"  (cost=13.33..17.40 rows=1 width=56)
                    ->  LockRows  (cost=13.33..17.39 rows=1 width=22)
                          ->  Bitmap Heap Scan on tasks tasks_1  (cost=13.33..17.38 rows=1 width=22)
                                Recheck Cond: ((state = 'CREATED'::nango_scheduler.task_states) OR (state = 'STARTED'::nango_scheduler.task_states) OR (state = 'STARTED'::nango_scheduler.task_states))
                                Filter: (((state = 'CREATED'::nango_scheduler.task_states) AND ((starts_after + ((created_to_started_timeout_secs)::double precision * '00:00:01'::interval)) < CURRENT_TIMESTAMP)) OR ((state = 'STARTED'::nango_scheduler.task_states) AND ((last_heartbeat_at + ((heartbeat_timeout_secs)::double precision * '00:00:01'::interval)) < CURRENT_TIMESTAMP)) OR ((state = 'STARTED'::nango_scheduler.task_states) AND ((last_state_transition_at + ((started_to_completed_timeout_secs)::double precision * '00:00:01'::interval)) < CURRENT_TIMESTAMP)))
                                ->  BitmapOr  (cost=13.33..13.33 rows=1 width=0)
                                      ->  Bitmap Index Scan on idx_tasks_state  (cost=0.00..4.44 rows=1 width=0)
                                            Index Cond: (state = 'CREATED'::nango_scheduler.task_states)
                                      ->  Bitmap Index Scan on idx_tasks_state  (cost=0.00..4.44 rows=1 width=0)
                                            Index Cond: (state = 'STARTED'::nango_scheduler.task_states)
                                      ->  Bitmap Index Scan on idx_tasks_state  (cost=0.00..4.44 rows=1 width=0)
                                            Index Cond: (state = 'STARTED'::nango_scheduler.task_states)
        ->  Index Scan using tasks_pkey on tasks  (cost=0.56..8.58 rows=1 width=296)
              Index Cond: (id = "ANY_subquery".id)

Query plan after

Update on tasks t  (cost=13.55..21.59 rows=1 width=123)
  CTE eligible_tasks
    ->  LockRows  (cost=8.89..12.99 rows=1 width=292)
          ->  Bitmap Heap Scan on tasks  (cost=8.89..12.98 rows=1 width=292)
                Recheck Cond: ((state = 'CREATED'::nango_scheduler.task_states) OR (state = 'STARTED'::nango_scheduler.task_states))
                Filter: (((state = 'CREATED'::nango_scheduler.task_states) AND ((starts_after + ((created_to_started_timeout_secs)::double precision * '00:00:01'::interval)) < CURRENT_TIMESTAMP)) OR ((state = 'STARTED'::nango_scheduler.task_states) AND (((last_heartbeat_at + ((heartbeat_timeout_secs)::double precision * '00:00:01'::interval)) < CURRENT_TIMESTAMP) OR ((last_state_transition_at + ((started_to_completed_timeout_secs)::double precision * '00:00:01'::interval)) < CURRENT_TIMESTAMP))))
                ->  BitmapOr  (cost=8.89..8.89 rows=1 width=0)
                      ->  Bitmap Index Scan on idx_tasks_state  (cost=0.00..4.44 rows=1 width=0)
                            Index Cond: (state = 'CREATED'::nango_scheduler.task_states)
                      ->  Bitmap Index Scan on idx_tasks_state  (cost=0.00..4.44 rows=1 width=0)
                            Index Cond: (state = 'STARTED'::nango_scheduler.task_states)
  ->  Nested Loop  (cost=0.56..8.60 rows=1 width=123)
        ->  CTE Scan on eligible_tasks e  (cost=0.00..0.02 rows=1 width=120)
        ->  Index Scan using tasks_pkey on tasks t  (cost=0.56..8.58 rows=1 width=22)
              Index Cond: (id = e.id)

I switched to use knex.raw because using knex syntax isn't shorter or easier to read I am also using a CTE instead of a IN subquery to make it slightly easier to read

Describe your changes

Issue ticket number and link

Checklist before requesting a review (skip if just adding/editing APIs & templates)

• I added tests, otherwise the reason is:
• I added observability, otherwise the reason is:
• I added analytics, otherwise the reason is:

[linear]

NAN-1774 Investigate orchestrator db spike and increasing latency