Oliver Schulz

I have this in #122 already, I think - if it's ok I'd hold off until that done (hopefully very soon) to reduce merge conflicts.

Thanks - I'll get back to that monster-PR, promise!

Included in #142

Instead of `pushfwd(f, finv, μ)`, we could do `pushfwd(setinverse(f, finv), μ)` if we implement JuliaMath/InverseFunctions.jl#18 . A `setinverse` construct convenient to have anyway, I think. Regarding with_logabsdet_jacobian - ChangesOfVariables can't...

Regarding the multiple call to `finv` - can we solve that with better forwarding mechanisms in MB somehow? Maybe we can avoid calling `insupport` separately, for example?

What if we specialize `logdensityof` directly as well for pushforward measures?

Could be nice. I remember running into of dispatch ambiguities with the sampler interface extensions in the past through, we'll probably need to do this carefully.

I'll add it to the ToDo-list for #122 ...

Oh, yes. Just an idea - how about using `Folds.reduce(op, executor = default_exec)(xs)` as a syntax instead, so that `Folds.reduce(op, executor = default_exec)` becomes a functional, semantically, just like `Folds.sum`?...

> It's a bit unclear to me in which scenarios this integration would be helpful. Naively, to me it seems if a package supports only ADType.jl then generally even with...