use GSL linear interpolator instead of cspline during initialisation in RichDet

@seaso Noticed an issue with the interpolators that was causing incorrect values to be returned. Part of the issue is, under certain circumstances the spline GSL interpolator used to initialise the home grown one I use was generating unstable values, if the input (x,y) values did not nicely describe a smooth function across the whole range of valid x values. This MR changes this to linear, which is not quite as good, but much more reliable as it cannot introduce such artefacts.

This MR will cause diffs in the nightly QM tests. I will update the Brunel refs in Brunel!201 (merged) once I have new versions from the nightlies themselves.