ENH: Improve type annotations and checking

[timmens]

Meta issue to collect ideas to improve type annotations and type checking:

• Use pandera to get more informative type-hints for pandas dataframes

  We use pandas dataframe internally (e.g., variable_info). It is very hard to check that a column exists or what dtype it has. We should discuss whether we should improve the type annotations for these objects, or replace the dataframes entirely by objects that are easier to type check and give a more descriptive structure.

• Use pydantic / beartype (or other tools) to validate user-input.

  We currently do a very basic validation of user input in each Model/Regime class. This simply duplicates behavior of better existing validation tools.