API should get Recommender dynamically

[frankcleary]

We'll need to get a recommender implementation for each API call, something like:

recommender = getRecommender(recommenders, otherContext)  # have a list of available recommenders
...

getRecommender(recommenders, otherContext)
   return random.choice(recommenders) # in the future can also use context to choose a recommender

[Freija]

I implemented this as follows:

• in app.py, a list of subclasses of the Recommendation class is constructed. One needs to import the subclass explicitly (like for example from recommender.gensimple import GenSimple) to make it part of that list.
• In the API call, the recommendation engine is chosen randomly from the list: this_recommender = random.choice(self.engine_classes)(corpus_name).