Add a method to generate records for Unit tests based on real data

[lfoppiano]

Previously discussed in #39, is possible to deserialize the java object of a record from the Nutch segment.

We could think this task in two phases:

1. We manually create the segments
2. We define an utility function that can automatically fish the right segment example, and store the specific example somewhere else (this may be checked in the repository)
3. Utility function that is called from the tests that deserialize the content Java object