fixed bug in historical forecasts not training model for first time if retrain is False (#1005)

Author: dennisbader

darts/models/forecasting/forecasting_model.py

@@ -435,7 +435,7 @@ def historical_forecasts(
 
             # train_cov = covariates.drop_after(pred_time) if covariates else None
 
-            if retrain:
+            if retrain or not self._fit_called:
                 self._fit_wrapper(
                     series=train,
                     past_covariates=past_covariates,