Feature/Model Testing During Training

This merge request contains following changes,

• Metrics class for performing default evaluation metrics (fedbiomed/common/metrics.py)
• Testing routine for TorchTrainingPlan and SkLearnSGDModel
• Processing testing argument in the Round class, to make sure is everthing OK to perform testing.
• Extending add_scalar message of HistoryMonitor, adding more information about real-time training/testing status
• json.py encode/decode MetricType defined in the experiment
• Changes in fedbiomed/researcher/monitor.py
  • MetricStore -> to store and categorize metric values received in real-time
  • Logging training/testing metric values to console as fancy/readable as possbile
  • Categorizing plots on the Tensorboard (with headers) based on metric name, metric for training/testing_global_updates/testing_localupdates
• Adding unittests for;
  • metric.py -> MetricTypes, Metrics
  • monitor.py -> _MetricStore
  • fedbiosklearn.py -> testing_routine
  • torchnn.py -> testing_routine