Fix dispatcher race condition when multiple scan jobs are dispatched in one go

[netomi]

This fixes #1734.

The dispatcher uses a transaction when processing queued scan jobs, which means that any change that is performed while dispatching the queued jobs will not be visible directly, but only when the transaction is committed.

Now it can happen that the dispatched jobs start right away while the dispatcher has not finished yet, but their status will still be queued instead of processing which results in a state where the job just returns but is not dispatched again.

This change should avoid that such that when a scan job is ready to be dispatched, the state is immediately changed to processing before enqueuing the respective job.

[netomi]

Additionally the method ExtensionScanCompletionService.checkCompletionSafely has been made Transactional as it calls a method in the same class that is Transactional to ensure that a transaction is really created.

[netomi]

@janbro