Support {VariationPath} for ntuple inputs

[alexander-held]

See use case described in #527 and some context in #435.

[MoAly98]

@alexander-held -- I am actually wondering why, in the first place, was this not supported? I think to start with a harmonised support to the overriding system independently of input type is a good idea.

[alexander-held]

This wasn't omitted on purpose: ntuple support came first, histogram support was added later. When implementing the first ntuple support I did not think of this yet. While designing the histogram interface it became clear that this is needed for that use case. #435 then captured some of the differences and possible things that may need to be added, which just had not been prioritized so far as I had not heard about concrete use cases prior to #527.