Posterior transform for objectives

[AdrianSosic]

Fixes #661 by properly handling Gaussianity-preserving (a.k.a. affine) transformations via a dedicated processing route utilizing BoTorch's PosteriorTransformation. Because the acquisition builder can now rely on this mechanism, much boilerplate could be removed from there (including the wrong logic causing #661) and be replaced with improved error handling instead.

[AVHopp]

@AdrianSosic For your reference, here is the reason that doc building is currently failing - seems like you need to adjust a link in the user guide