In RelationalSnapshotChangeEventSource, inside createDataEvents, we queue up the IN_PROGRESS notification for each snapshotted table before submitting the job to the executor. If there are more than enough snapshot threads for all collections, this doesn't pose much of an issue, but in the event that the number of threads is less than the number of collections, the notifications are sent well before the exporting of that table's data begins.

We should move the call to notifyTableInProgress into the Callable which is defined based on createDataEventsForTableCallable so that the notification is sent only when the task begins execution.

This will then mirror the same behavior that MongoDB uses.