Implementing the Run3 FT algorithms

Implementing the Run3 FT algorithms Requieres upstream changes of Rec!4638 (merged) and LHCb!5376 (merged). Requires files provided by lhcb-datapkg/ParamFiles!164 (merged)

The new Run3 taggers are composed of a configurable track selection and a NN-based mistag estimation. The tag decision is then made by the charge of the track providing the best (lowest) mistag estimate among all tracks passing the selection. This is implemented in Phys/DaVinci/python/DaVinci/FlavourTagging/onnx_flavourTagging.py, where the https://gitlab.cern.ch/lhcb/Rec/-/blob/72c7fc4899eb90a4e9627cb8c666051d58527e00/Phys/SelAlgorithms/src/MLServiceAlg.cpp is used for the NN inference. The NN input features, the corresponding ordering, the selection, and the definition of composit functors are defined in string format and provided in json format lhcb-datapkg/ParamFiles!164 (merged). Here, also the NN models are stored in onnx format. Integration tests for these taggers are implemented in LHCbIntegrationTests!111 (merged), with the necessary options files for the tuple processing provided here. This goes together with the addition of a version enconding of the previously ported Run2 taggers (Rec!4638 (merged)) and the corresponding clean up of the tagger namespace (LHCb!5376 (merged)).