implement varx by Lucas Parra

[behinger]

I just stumbled over this awesome paper:
https://arxiv.org/pdf/2404.10834
with code here: https://github.com/lcparra/varx (check license)

I think we should explore this method - it looks really cool :)

[behinger]

if I understood correctly - this should be straight forward:
simply add a 1-lagged continuous TRF predictor to the model.

This will be much more interesting, once we got group-lasso integrated somehow... :S