WIP:Implementation of CARL Neural Network Reweighting PMG Tool

Purpose

The following merge request adds a new Open Neural Network Exchange using the CARL neural network re-weighting technique outlined in the QT task AGENE-1726.

Features

The PMGTools:: namespace is to be expanded with an additional CARLNN class that reads, and builds in memory a PyTorch trained neural network (see carl-torch). This neural network is stored using the new Onnx file format, which is run within a C++ API using OnnxRunTime.

The CARLNN:: is to be implemented by Athena Derivation jobs as a neural network reweighter that will add additional internal MC Event Weights to the xAOD::MetaData/EventInfo for DxAOD disk storage.