Scheduled HangFire Jobs duplicate on server restart

If you use an activity like `StartAt`, then Elsa will schedule the `ResumeWorkflowJob.ExecuteAsync` job in Hangfire to run at the specified start at timestamp.
Hangfire stores those persistently in the Hangfire.Jobs table, so they survive a server reboot.
Everytime the Elsa server application rebootst though, Elsa will schedule another `ResumeWorkflowJob.ExecuteAsync` for all workflow instances currently suspended on such an activity, regardless if a job is already scheduled.
This very rapidly creates thousands of duplicate scheduled jobs in the hangfire dashboard.

Seen in Elsa 3.6-rc2

Start by tracing where Elsa schedules ResumeWorkflowJob.ExecuteAsync for workflow instances suspended on a StartAt activity, then inspect how existing scheduled jobs are represented in Hangfire.Jobs across a server restart. Done means restarting the Elsa server does not create duplicate scheduled jobs for the same suspended workflow.

Written by the indexing model from the issue text.