Add support for noisy optimization

[danielhorn]

An optimization path should know wether the optimization is noisy or deterministic.

The main difference is: In deterministic optimization, there will be a single objective value / vector for each parameter setting. In noisy optimization, there can and will be multiple different objective values. This should be represented be our internal structure.

Moreover, there should be some specialized getter-Functions for noisy optimization, for example:
getDistinctOptPathLength - how many different settings have been evaluated?
getNumerOfEvaluations - for each parameter setting: how many evaluations have been done
getAggregatedOptPath - use an aggregation function on the objective values for each parameter setting. Most common application: Get the mean value.

Some existing funktion must be adapted. For example:
getOptPathParetoFront and getOptPathBestIndex should return the best Elements value with respect to some aggregation, for example the mean.

We must think about how to extend the concept DOB and EOL. Are they properties of a parameter setting or of an evaluation?

I will do this in the near future, but we should talk about the details before.