uf_addmarginal()/ uf_predictContinuous() doesn't warn if covariate out of bounds

[ReneSkukies]

Using uf_addmarginal(uf_predictContinuous(ufresult,'predictAt',predictAt)) currently, silently fills predicted effects with zeros when trying to predict out of bound values of a covariate (e.g. marginalizing on reaction time of 250ms of a subject with the fastest reaction time of 290ms).

This should at least throw a warning to the user.

[olafdimigen]

Important, indeed, and I would not be surprised if this sometimes happended with (very small/large) saccade sizes in the past.

Unrelated: I wonder whether we should implement cross-version unit tests (comparing unfold.jl vs Matlab-unfold results for the same data and model) to automatically spot issues and/or inconsistencies.

[behinger]

I'm not super happy with uf_addmarginal/uf_predictContinuous anyway; but I didnt know better at that time. A warning should be super simple but very helpful. I'm not sure I can really do it at the moment, I could point to the right direction (line of code + what roughly should work) if needed though

Regarding cross-version tests: indeed; but I dont think this one would have been caught, it is too detailed. It could rather capture other inconsistencies (which surely exist)