Feature request: carry over model

Feature Request: Add Carryover Effect Modeling Similar to LightweightMMM

Currently, Medrian does not support carryover models like those implemented in [LightweightMMM](https://github.com/google/lightweight_mmm). This limits its ability to handle cases where the effect of a marketing channel is delayed and reaches its peak impact after a certain lag.

It would be very helpful if you could add support for modeling such carryover effects. This would significantly improve the applicability of Medrian in real-world scenarios where marketing efforts do not have an immediate effect.

Thank you for considering this feature!