Implement an algorithm for photon/pi0 discrimination based on input features and a trained model in ONNX format

[giovannimarchiori]

The code in this PR provides a Gaudi algorithm that

• reads the list of input features from a JSON file and the trained BDT model from an ONNX file
• for each cluster, reads the input features from the shapeParameters vector (if available), runs the inference, and saves the photon probability as an additional shape parameter of the new output cluster collection
  It works with ONNX files created with either XGBoost and LightGBM (tested with both).
  Could probably work also with different models based on features (DNN..)

Uses as inputs the shower shapes calculated in #79

[giovannimarchiori]

Just rebased..

[giovannimarchiori]

Hi @kjvbrt , I implemented the handling of exceptions in reading the JSON and rebased - is there anything left for me to do?

Thx, Giovanni

[kjvbrt]

Hi @giovannimarchiori, nice :)
Merging...