Write a script testing the relevancy of the current recommendation algorithm allowing also to be easily configured to test any new algorithms implemented.

The script should be given one parameter: the reference to the guess rating field in Rating to be used (for the current algorithm: guess_rating_alg1)
The script should simply compute the average difference between the real rating (the one manually assigned by the user) and the guessed rating, i.e. sum(abs(rating-guess_rating)) / number_of_ratings_with_guess_rating_computed

It should be possible with one PSQL query.

When implementing new algorithms we'll be keeping the old ones and only adding new fields:
- new score fields in RatingComparator (current algorithm: score - to be renamed to score_alg1, new ones to follow the schema)
- new guess rating fields in Rating (guess_rating_alg2, guess_rating_alg3, etc)