Remove scaling by WeightNormalisation in Sherpa_i
The arbitrary scaling of the weights by WeightNormalisation means that samples with different biasing but otherwise equal phase-space coverage cannot easily be combined at EVNT level (or not at all at DAOD level, depending on how the sum-of-weights is propagated).
This change set removes this scaling, which will give more flexibility for combining specialised MC setups with the bulk samples.