Add score based test evaluation

[orangejulius]

Roughly, this evaluates tests in a more granular way than simply everything passed vs anything failed as in most unit-test style suites. See the original fuzzy tests problem statement(pelias/acceptance-tests#109) for more info.

Here are some of my notes on the actual implementation details, mostly written up to get a grasp on the low level task at hand:

• how to score a single test
  • situation:
    • many api results (as js objects with properties) returned from query
    • many expectations specified in tests (as js objects with properties)
    • both js objects potentially contain other objects nested
    • the order of the expectations and the corresponding result returned are not related
    • the basic task is to return a score, both actual and maximum, for each expectation and in turn the entire test
    • the most common situation would assign one point to each matching element within an object
      • however, additionally, different fields should be able to be given different weights to adjust for their importance
    • some expectations, instead of listing an object with properties, simply give a string
      • this string is used to look up an actual expectation object in locations.json
      • if a matching object is found, use it instead
      • if not, the test is marked as a placeholder. the score for the test is not relevant but continuing to compute it might be ok
    • determining pass/fail for the test
      • the test passed if the score is equal to the highest possible score (all expectations match), and the test is not a placeholder