[Feature Request]: Integrate Clusters into the DoubleMLData Class

[SvenKlaassen]

Describe the feature you want to propose or implement

Clusters should be element of the basic DoubleMLData class.

Propose a possible solution or implementation

No response

Did you consider alternatives to the proposed solution. If yes, please describe

No response

Comments, context or references

No response

[JanTeichertKluge]

• Create custom data backend/classes for model-submodules.
  • The base data backend should include $Y$, $D$, $X$ and instruments $Z$.
  • Then, e.g. t is only added to the DID data Backend, s only for sample-selection models, etc.
• n-obs needed for variance estimation (for panel data, n_ids = total observations, scaled by sqrt n * t). May need different n_obs for different asymptotics.
• If the Base Data class has implemented the cluster options, how should the workflow, whether ClusterData or not, be handled? By flag indicators or by dynamic properties of the dataframe (cluster_id = None or similar?)