Now our neural network can make stable classification on 3 algorithms with 50% accuracy

Added cross-validation, graph parallelism degree metric introduced

…h-Lab/sampo into feature/nn

…tall 'sklearn'

Now the neural network is well able to predict the most suitable scheduler for a certain work schedule.

Additionally, there is an explanation of neural network - code that uses Lime library can describe the behavior of the net.

# Conflicts:
#	pyproject.toml
#	sampo/scheduler/generic.py
#	sampo/scheduler/timeline/just_in_time_timeline.py
#	sampo/scheduler/timeline/momentum_timeline.py

# Conflicts:
#	pyproject.toml
#	sampo/scheduler/generic.py
#	sampo/scheduler/timeline/just_in_time_timeline.py
#	sampo/scheduler/timeline/momentum_timeline.py