Model comparision should report what datasets are used for comparision

[javier-alvarez]

It is non transparent if the datasets are exactly the same version or we are just comparing patient ids.

More debugging info in general would be good.

AB#3910