improve hash algorithm

[michielrogissart-cognizone]

For now, the hash function works as follow:

• if the model does not contain blank nodes
  • compute hash of statements as string
  • XOR the hashes
• if the model contains blank nodes:
  • create sorted turtle
  • compute hash of sorted turtle

This is because there is no standard way to assign IDs to blank nodes. To ensure that the algorithm always returns the same hash for the same model, these ids shouldn't affect the hash.

Note that the sorted turtle would also work for models without blank nodes, but it is faster to not sort the statements.

A better approach would be to have predictable ids for blank nodes so that the first method can be used for models with blank nodes.