Closes #597

This is required to run the tests.

Just reset the current database selected.

This is blocking me to upgrade Rails to use thor 1.4.0.

None of the Rails dependencies have upper bounds on any gem
to avoid this kind of issues.

Fix the devcontainer configuration

Even though this will be overwritten by Active Job when called through
`perform_later`.

This is necessary for recurring tasks, where we run the `enqueue`
callbacks manually and call `Job.enqueue`. We rely on `successfully_enqueued?`
to know whether we need to record the recurring execution. We were setting
that no matter what after calling `Job.enqueue` because we were relying on
that to raise (which would be the case for Active Record errors).

In the case of discarding it because of concurrency controls, the job is simply
not persisted but no error is raised, which means we were trying to
record a recurring execution without a persisted job, which would raise.

Fixes #598

Especially with SQLite, where the forked process claims the job before
the main process.

See #602

See #602

Don't try to enforce sequential order for jobs, given that the order is
not even guaranteed by Solid Queue. Rename the test job to reflect that.
Fix also another test where we waited for a job to finish but not for it
to release the semaphore, so it happened most of the times but sometimes
it didn't.