E=1 path for energy score losses could cause issues

[azrael417]

For the E=1 case, we should guard the EnergyScoreLosses. Ideally by setting the spread term to zero, since am empty sum is zero. We do the same for the CRPS, would be good to mirror that behavior.