Collect dataset for ML

[ryukzak]

That task includes:

• Defining dataset format in accordance with @iburakov requirements.
• Traverse over all known synthesis tree nodes and store data into memory (maybe a bad idea due to synthesis tree size) or output data stream (sounds more adequate, maybe it is the lazy list is applicable here).
• Synthesis tree exploration method for finding all necessary nodes.
  • maximum deep should be restricted.
  • ignore rule
• Select and implement one of the following API (need to coordinate with me):
  • CLI and stdout
  • CLI and file
  • REST API

[iburakov]

Example of input data used in .ipynb prototype:

counter.lua.csv

• alt_bindings, alt_refactorings, alt_dataflows - alternative bindings, refactorings and dataflows for a node (siblings).
• pOutputNumber,pAlternative,pWave,pAllowDataFlow,pCritical,pPercentOfBindedInputs,pPossibleDeadlock,pNumberOfBindedFunctions,pRestless,pNotTransferableInputs,pRestrictedTime,pWaitTime,pRefactoringType - available decision metrics, null if not available.
• label - calculated label for a node (model training target). See the prototype for calculation algorithm.

[ryukzak]

Closed in #167