Added ensemble support for RingCounting tool.

[s4294967296]

• Moved logic for processing predictions to method process_predictions to improve clarity
• Added new configuration variables model_is_ensemble, ensemble_model_count,ensemble_prediction_combination_mode
• Added support for loading ensemble models and processing predictions
• Added support for average and majority-voting ensemble
• Added storing of majority-voting ensemble class prediction in new variables RingCountingVoting{SR,MR}Prediction in RecoEvent BoostStore. The averaged predictions (currently preferred mode) will simply be stored in the already existing RingCounting{SR,MR}Prediction BoostStores.
• Updated documentation within RingCounting.py
• Updated README.md
• Updated configfiles/RingCounting/RingCountingConfig

Users using the RingCounting tool need to update their RingCountingConfig file(s) following this change.

An improved classifier consisting of ~10 CNN models will soon be available.

[brichards64]

Hi Daniel,

can you turn up to the software meeting to walk us through this change. Or if you cant make the software meeting lets meet offline at some point, so i can merge this in.

Ben