hubEvals 0.2.0

• Add support for scoring sample output types via transform_sample_model_out() and the "sample" case in score_model_out(). Marginal scoring produces metrics such as CRPS, bias, and DSS; compound scoring (via the new compound_taskid_set argument) produces multivariate scores such as energy score and variogram score for joint forecasts (#94).

• score_model_out() now errors with a clear message when a scale transformation produces non-finite values (NaN or Inf), instead of silently returning invalid scores (#99).