Framework implementation

• ONNXRuntime ultimately failed with compiler directives (asm volatile("s_mov_b32 m0, %0; \n\t") -> Chat with Giulio and Sergio: This is probably due to some incompatibility with our gcc and might only be fixable with gcc 14
• Trying new versions of PyTorch again
• Nicodemos is fully up to speed and trying new potential framework candidates: Tensorflow lite -> Light inference framework of TF made for deployment on mobile devices and various accelerators

NN implementation in O2

• NN's are working in the GPU clusterization code, but some issues still to be fixed (e.g. tpc-native-clusters.root is not being published correctly, to be investigated)
• Added some options (like model paths, thread settings, etc.) to GPUSettingsList.h which makes the workflow more configurable
• Built a track matcher macro which compares two tracks and their associated clusters to compare if they can be counted as the same or not (here it woud be good to sit together and decide on some criteria)