forecast bootstraps

• use data from #49 (closed)
• create forecasts with keras model
• save bootstrap forecasts (?, evaluate if required or not)