Changes to the MapReduce algorithm.

[GoldenBadger]

As discussed in the Sprint 5 planning meeting, we would like to rework the mapreduce algorithm itself to be closer to Apache Hadoop's flow.

The proposed order of operations is as follows:

1. Mappers get data from external data source (FS).
2. When a mapper completes all of its assigned tasks, it performs a Combine step followed by a Partition step on the intermediate data saved to its local disk.
3. When all mappers have completed their tasks, the Shuffle step begins, which is simply the movement of local intermediate files to the FS for further processing.
4. Then the Merge step combines the values of all duplicate keys.
5. Finally, the intermediate data is sent to the reducers for final processing.