[DOC] sort slightly disordered forecasting API reference

[fkiraly]

The estimators in the forecasting API reference were slightly disordered, some estimators were sorted in the wrong category.
The mismatches were mostly statsforecast based estimators which were mistakenly added below the first such model - in the ARIMA category, even though they didn't belong there.

I made the following changes:

• sorted the statsforecast models in the right categories, Auto-Theta, Auto-ETS and -CES need to go in the exponential smoothing category
• split the AR/MA category into plain (V)AR(I)MAX and Auto-ARIMA models
• added an intermittent time series category