Add some type checks to algorithm/estimation/estimators.py

[callahanr]

Here's get_primal_point_estimate:

def get_primal_ipw_point_estimate(
    *,
    data: pd.DataFrame,
    graph: NxMixedGraph,
    treatment: Variable,
    treatment_value: int | float,
    outcome: Variable,
) -> float:
    """Estimate the counterfactual mean E[Y(t)] with the Primal IPW estimator on p-fixable graphs."""
    # TODO: This function currently returns type Any to conform with mypy requirements. 
    #       That's not the best reason to make a return type less restrictive. 
    #       Consider warning the user if the type of the return value cannot be 
    #       coerced to a float. -callahanr
    beta_primal = get_beta_primal(
        data=data,
        graph=graph,
        treatment=treatment,
        treatment_value=treatment_value,
        outcome=outcome,
    )
    return np.mean(beta_primal).item()

This function currently returns an object of type Any because beta_primal is of type np.ndarray[Any, np.dtype[Any]]. Currently to conform to mypy standards we allow the return value to be either a float or Any, but ideally we would want to coerce the return value from Any to a float if possible, and if not then warn the user that something may have gone wrong with the computation.