Small improvements to fitness functions and evolutionary algorithm

[lisv]

Following the meeting on the 19th of April:

• There seems to be an issue with the agglomerative clustering due to the depth of the json object not being handled correctly: things inside arrays/jsonobjects in objects were not emptied when creating askeleton. Therefore many individuals create their own agglomerative clustering object because the response skeleton is different for each individual.
• Selection is always done using elitist selection. Tournament selection may work better for the purpose of creating diverse requests.

[lisv]

Update

• Fixed the agglomerative clustering strip values method
• Implemented tournament selection