Modelling group-level variance of random slope/coefficient

It sounds like maybe you are asking something similar as on this post?
Maybe the suggestion of the LMMELSM package on that thread would help you?