Implement Angle Statistics Reconstruction (ASR)

[GilesStrong]

Stapleton et al 2014 introduces an inference method they claim is better than PoCA and less complex than MLEM. The main benefit seems to be that the point assigned to the scattering is constrained to lie within the volume.

Given our voxel-wise integration of 3D Gaussians centred on the PoCA, this seems to be less of a problem for us, but still the algorithm may bring additional benefits.

It is also unclear whether the algo. can be used directly for detector optimisation, due to the hard assignement of predictions to voxels, however the algo. computes weights per voxel based on their transverse distance to the muon trajectories, so this could be used instead to smear the X0 predictions across the volume, based on their closeness to that predicted path of the muon.

@MaximeLagrange is investigating and can comment further.