This PR is part one of a replacement for #350. It introduces data set structures for pure substances and binary mixtures. The goal is to make working with (experimental) data collections easier.

Data sets are provided for pure substances and binary mixtures. A data set is connected to a property such as vapor pressures or bubble point pressures. Once created, it can be evaluated from currently existing equations of state (non-AD). Furthermore, for models that support parameter-ad, parameter gradients can be computed. In a next PR this will be connected to a regression framework.

Infrastructure changes:

• I made ndarray non-optional. Can be reverted if there are strong opinions about it.
• Calculations for properties moved to own files (was in ad/mod.rs)
• I tried making ParameterAD more robust also with respect to the regression framework. Could be more elegant but a better way would be to actually redesign our parameter handling in general so for now it should do the job.