Forecast future values of a univariate time series.

Args: values: Historical observations in chronological order (minimum 4 points). horizon: Number of future steps to forecast. seasonal_periods: Length of one seasonal cycle (e.g. 12 for monthly data with yearly seasonality), if the series is seasonal. Omit if unknown or the series is too short to estimate seasonality reliably.

Returns: Point forecast plus an approximate 80% confidence interval, and the method actually used (seasonal models silently fall back to a trend-only model if there isn't enough data for the requested period).