Implementation of Experiment Framework (Multiple Configurations + Scheduler + Metrics)

[cassianobecker]

The .ini file was out of data, updating it. Please run config.py as main to run a test.

[cassianobecker]

This pull request includes the following improvements:

experiment framework:

• handles multiple .ini configurations and performs cartesian product expansion
• top-level experiment scheduler runs (parallel) experiments in debug, local, or cbica(gpu) modes
• extensible metrics framework to collect computational, algorithmic, and statistical data from experiments
• clean experiment template with configuration of metrics collection through the metrics.ini file
• metrics renderers that save results to text files and/or numpy array files, according to configuration in observers.ini

pre-processing:

• separated pre-processing code from dataloader
• integrated with Linden's implementation of diffusion masks

other:

• unification of dataset (download and postprocessing data) and results outputs to standard folder ~/.dnn created automatically
• automatic detection of cbica environment (no need to change .ini file)