Remove unneeded include of pg_locks in query when displaying jobs table

[jgrau]

Related to #1540

• Removes an (unused?) join of the pg_locks table which was slow

@bensheldon regarding your comments

• The pg_locks could be replaced with a locked_at IS NOT NULL query

I'm not sure but I found that the joined table was not being used for constraints but for including the locktype which seems to be unused(?). I removed it and browsed around the dashboard and found to errors 🤷

• The COALESCE can be replaced solely with good_jobs.scheduled_at because now all jobs should have a scheduled_at regardless of whether they are expect to run immediately or in the future.

Can you elaborate? There's multiple places where we use coalesce(scheduled_at, created_at) so if that's not needed in general anymore I feel there's a lot of cleanup that could be done and I can do that in this PR with your guidance :)

[bensheldon]

Yep, the COALESCENCE is not needed. Prior to GoodJob v4 the scheduled_at column would be null for non-future scheduled jobs. That's no longer the case as of GoodJob v4 (some discussion in #1323 (comment))

I think removing the coalesce might be a slight behavior change in some cases, but I think it might be better to be explicit about whether we intend to order by when the job was first enqueued (created_at) or when the job is intended to be executed even when retried (scheduled_at)

[bensheldon]

Maybe let's do the COALESCE removal in a separate PR so I can land this one sooner. I want to poke at it a little, but this looks good 👍