It was difficult to understand but easy to implement :-)
Some mechanisms has been added to avoid to duplicate
quiz_dataset_definitions (confirmed from Henrik).
Please, test, test, test, test....
If it exists in backup-file used, else, old search-by-name is used.
When inserting new categories, stamp is filled it it doesn't exist.
Updated version to 1.1 Beta +++ !!
The process look for category, stamp and version to calculate
when to create a question or use a existing one.
This avoid duplicates in RANDOM and RANDOMSAMATCH qtypes.
