[ENH] Add PatchTST model to v2 interface

[Sylver-Icy]

Currently, PatchTST (from the THUML Time-Series-Library) is not available in pytorch-forecasting. Adding it would expand the library with a modern transformer-based architecture that has shown strong performance in long-term time-series forecasting benchmarks.

I would like to implement PatchTST using the new _base_model_v2.BaseModel interface.

I have checked the repository and open PRs and do not see anyone currently working on PatchTST.

Migrating an existing v1 model (e.g., NBeats) to native v2 was considered, but those are already being worked on. Adding a new model avoids duplication and expands the model collection.

If approved, I will proceed by:

1. Creating the model under pytorch_forecasting/models/patchtst/
2. Implementing it natively on top of the v2 BaseModel
3. Adding minimal tests and documentation

[amruth6002]

hii @Sylver-Icy , i have already raised this issue #2067 and submitted a PR #2099.