Decouple spline tables creation from trial generation

[mlincett]

It was observed while investigating the (now closed) #181 issue that the make_SoB_splines.py routine (and in general the creation of spline tables for the dataset) is not decoupled from the trial generation.

When an analysis is first tested interactively and then submitted to the grid (assuming a share filesystem), then the cluster jobs will use the previously created spline files. If the spline files do not exist, then the jobs will try to create these files all at once, which is not very efficient.