Replace Scala parallel collections with streams

[gaurav]

There are two main things that need to be done here:

• In my current quick-and-dirty setup, RoboCORD loads all JSON files that it needs to use into memory. This should be replaced by a lazy system so that the file isn't actually read and parsed until it is needed.
  • We use a pretty simple Scala parallel collection system to process documents in parallel. This seems to work well enough for now, with jobs usually taking <15m to process 105 articles each, but eventually we might want to replace this with something fancier.

[balhoff]

In my current quick-and-dirty setup, RoboCORD loads all JSON files that it needs to use into memory. This should be replaced by a lazy system so that the file isn't actually read and parsed until it is needed.

Streams!