Flow training in unconstrained space

[michaeldeistler]

This should happen only after we have moved to pyroflows

[janfb]

addition: would be great to add log transforms for bounded data, e.g., for reaction time data in NLE.

[michaeldeistler]

Removing the label because we will do this after the flowtorch port

[janfb]

More context: With #952 #957 getting fixed, it will be possible to add new density estimators, e.g., those from zuko.

Goal: some priors are bounded, but it can be beneficial to perform the density estimation in unconstrained space, e.g., add a transform that transforms the parameters to an unconstrained space before or after the flow.

[bkmi]

I literally just implemented this for another project. It shouldn't be too hard.

[bkmi]

actually I will work on something else for now.