5001-node-selection-algorithm speed increase.

[burdettadam]

20 minutes to less than 10 seconds on my test data.
To get this speed increase:

• I removed unique_combinations and used combinations from itertools.
• I utilized Pool from multiprocessing to do a MapReduce on batches from unique_combinations.
• I removed dead code and reorganized some things to reduce calculations.
  I do not have time to work on this code anymore. If needed, more speed could be achieved by using panda to read in the config and create a table that would allow unique combinations to be represented by indexes of that table. this would cut down some un-needed computation. ComboAnalysis could also be optimized by having a pre-allocated array for maintaining the best combinations. I saw significant time being spent in an array append method inside ComboAnalysis, with the way batches are being processed, this would be a significant speed increase.
  Another possible speed increase would be using a library like numpy for all math operations.
  This code needs to be tested more than what I have done.