Enable transformation of list attributes to average and confidence interval

[thomaskeller79]

Currently, list attributes can only be reported by listing all elements. In this issue, we'd like to introduce the possibility to transform a list attribute into an attribute that consists of the average and a confidence interval, e.g., to turn the round-reward attribute into the average-reward and a confidence interval.

This issue was started as prost-issue-100 of the Prost planner repo and has been moved to prostlab. The discussion there ended with the insight that the studentized bootstrap and the empirical bootstrap method lead to vastly different results and that it is not clear which is the right one to support. If there is no clear answer, we should implement both.

[thomaskeller79]

The implementation we have currently adds a filter to the report generation step which generates the confidence intervals from the raw data. Instead, we should implement this in the same way as we did with the IPC score, which is computed during parsing and added to properties directly such that report generation just needs to read the data.