[ENH] Wishlist of forecasters and integrations

[fkiraly]

We should start adding models to the v2 interface as designed (possibly with predict adaptation later on).

A good starting point is to look at the implementations of DLinear, SAMformer, and TFT in the v2 implementation.

The "recipe" is trying to add a new estimator which is compliant with the v2 interface. For each model, it should also be decided whether interfacing (import and dependency management) or forking (copy and adapt) is better.

Layers go in layers, the models go in models, and models import layers from layers.

Contributors should pick one model and post about it here:

List of models to look at:

• current models in pytorch-forecasting that do not have a v2 implementation
• the DSIPTS package https://github.com/DSIP-FBK/DSIPTS
• the thuml package https://github.com/thuml/Time-Series-Library
• forecasting foundation models from this list [ENH] umbrella issue - foundation models and pre-trained models sktime#6177 - note that this is difficult until weight handling is designed in v2, but API design work is appreciated!

Also, anyone should suggest models below to add to this wishlist.