Use GSR phrase info to improve UP

[ahalterman]

The Spanish and Arabic GSRs both have information on which words in the sentence correspond to the source actor and target actor. This is very useful because it can be used to test whether UniversalPetrarch's rules for extracting these phrases are good or not, without having to depend on the quality of the dictionaries. We should have a test suite that looks just at UP's ability to recover these phrases across languages.