PG + Affine + Residual model

[balbasty]

The pgra_model script allows to build a model with PG + Affine + Residual

• in PG: the scaling step is still a work in progress.
• in PGRA: optimisation of the residual regularisation is deactivated for now.
• in Affine: For now, a prior is optimised over all affine parameters (including translation/rotation). This will change later.
• In batchProcess: I have started writing a distribute function to deal with the cluster. I wasn't able to test it so it likely does not work.
• Generally: I'll need to add some variabl cleaning. There is no need to keep that many values in the final model.
• Template update: I need to use a smooth prior for the template update. For now, I am doing either ML or ML + gaussian smoothing.