Why? current way of developing nhp_model_databricks alongside nhp_model is a bit convoluted
If it works, we could also pip install in the notebook the specific version of the model from the wheel saved on Azure. Then we would not have to specify a specific version of nhp_model as a dependency in the pyproject.toml. This means that we have fewer places to change the model version when deploying a new version of the model