Column constructor

[MarcoGorelli]

For constructing DataFrames, we have DataFrame.from_dict

How about constructing a Column? Example of where this would be needed:

https://github.com/mwaskom/seaborn/blob/5d9f37159bbd3ac44c8c8a06825583ba25648525/seaborn/_oldcore.py#L1217-L1220

        for var in axis_variables:
            other_var = {"x": "y", "y": "x"}[var]

            converter = pd.Series(index=self.plot_data.index, name=var, dtype=object)

[rgommers]

Seems like we need this. One comment during today's call was that these kind of constructors are quite messy in existing libraries, mainly for two reasons: they take a large number of types of input (arrays, scalars, sequences, generators, etc.) as well as do dtype inference on the contents. It looks like we want to avoid that, by only accepting well-defined input types (probably sequence and array?) and having the caller specify the dtype. Each element of input sequences must be of that dtype.