Improvement: Add ability to forecast with prediction intervals

[petersonR]

I'm not currently sure the best way to do this, whether via simulations or otherwise, but it would be nice to obtain forecasts and prediction intervals for fitted fastTS objects.