Job runs/enqueued multiple times if server restarts while job is being processed

Scenario:
- Configure the server to use sliding visibility
```csharp
SlidingInvisibilityTimeout = TimeSpan.FromMinutes(5.0),
QueuePollInterval = TimeSpan.Zero,
```
- Enqueue a job
- wait for it to be in the processing state
- restart the job server while the job is being processed
- The dashboard shows a notification that the server stopped resopnding
- click the re-enqueue button in the dashboard for that job before the job is re-enqueued by the system itself

Actual result:
- After the sliding visibility timeout is exceeded, a second instance of the job starts.
- That's because the JobQueue table gets another entry for the job and the legacy one is used as well when the FetchedAt timestamp is outrun by the timeout

Expected result:
- A job can not run multiple times concurrently
- There is not more than one entry for each job in the JobQueue table, even if the server restarts/crashes etc.

Reproduce the reported scenario with SlidingInvisibilityTimeout set to five minutes and QueuePollInterval set to zero, then inspect the JobQueue table across a server restart and dashboard re-enqueue. Confirm how the legacy and newly created entries are consumed after FetchedAt expires. Done means a restarted or crashed server cannot leave duplicate JobQueue entries or run the same job concurrently.

Written by the indexing model from the issue text.