Currently `pd.DataFrame.astype` only allows for one type. I would like it to support a `dict` mapping column names to dtypes. Like for example the `dtype` argument of `pd.read_json`. ``` python df = pd.DataFrame({'a': [1, 2, 3], 'b': [1, 2, 3]}) df = df.astype({'a': 'int16', 'b': 'float32'}) ```