Integration Tests and Unit Tests should use random KeyData

[ubamrein]

Currently, in the unit and integration tests we use fixed, hardcoded keydata. Since keyData needs to be unique in the database, we introduce potential problems within one test class, when we make assertions on the number of keys in the database (e.g. two times a key with data "a" gets inserted, the second overwrites the first).

We propose to add a utility class, which provides a random keydata.