Added momentum vector estimation to neural network output

• Projections from last week were correct, but tracks are evaluated at a clusterX which is not within the TPC (red points: track evaluated at X of a given cluster associated to a track), even though the clusterX seems to be fine (blue points all within the TPC) 
  
  
• Used track momenta anyway for now to assess NN performance:

Axis is in [GeV/c]. Not exactly sure why there are these bumps in the X-momentum distribution. But seemingly got a first reasonable momentum vector estimation working!

Position resolution decreased. Previously std_dev ~= 0.11, now 0.18. Probably need bigger network for the fit (very small so far, 2 layers)

Sigma estimation is shifted

Best idea so far is to increase the network size and redo fit. Momentum vector estimation can now be made, techincally also on real data!