add feature importance for statsforecast models that accept exogenous features

[GGA-PERSO]

Description

As every time we use Exogenous features we need something to explain a posteriori why a feature brings more value than an other one to the final prediction.
Feature importance or SHAP are some tools that we have in MLForecasts
I would love to have somehing similar in Statsforecast, specially for models that accepts exogenous features (like mfles for example)

Use case

I want to add several lags of a feature. I don't know which one brought more value to my final prediction

[sm-ak-r33]

Following