Improve error message when future covariates are missing during backtest() call

Is your feature request related to a current problem? Please describe. Yes.When using backtest() with a model trained with train and val future covariates, forgetting to pass the full future covariates again at backtest time results in an error (IndexError: index 1 is out of bounds for axis 0 with size 1.).

Describe proposed solution A better explanation or error message should be thrown as it is the case for "regular" historical forecasts.

Additional context Error currently thrown :

IndexError: index 1 is out of bounds for axis 0 with size 1.