Should we support multiple xnat configuration sets per dataset

[mih]

Right now the configuration names look like:

datalad.xnat.NAME.authentication_type

where NAME defaults to `default, but could theoretically be different.

However, this kind of scope setting is not accessible. Keep it? Rip it out?

[aqw]

From my view, this could be useful in a situation where a facility maintains two XNAT instances, one for public data and the other for internal/private/protected data. It would allow datasets to be organized in ways that don't necessarily have to follow/mirror the model that the infrastructure is deployed.