Fix #34.
Make the benchmarking procedure more efficient In cases where the system wants to only be tested in data loading. In particular:
- There is now no need to load the generated data to compute the number of added, deleted and total triples in order to later compute the KPIs related to bulk loading
- Versions construction as independent copies distinguished from data loading (for computing the expected results)
- In DG load the data and compute the expected answers if and only if at least one query type is enabled