Ranking per slice

Current ranking implementation takes the last observation into consideration. This enhancement should allow different ranking strategies by taking advantage of the data cube properties. For example rank different versions of the same dataset on multiple metrics
