Slowly getting closer to the new design. Observations and StateVector are now class based, but objects are getting more elaborate each time. One remaining problem is the relation between the StateVector, the ensemble members, and the observations. These are all linked but no object of one type can easily contain the other. 

For instance, does a model mixing ratio sample belong to an observation object because that is what it represents and is derived from? Or is it part of an ensemble member object because each member creates one sample set? And does a member object belong to the state vector and if so, how does it play together with the ohter time steps in the filter? Are these all separate StateVectoir objects or is there only one StateVector with nlag sets of data??

The issue with nlag is not nicely resolved now.

Next task is to build the optimizer such that it takes all info from the different objects and performs a LSQM.