How is it that the SeasonalNaive forecaster takes a 'level' argument in its predict method? I thought seasonal naive just takes the value from the last cycle which it finds by looking at the season_length. How can it calculate confidence intervals? How does 'level' work for it?