Improve typing when extracting values from pylibcudf columns

A pylibcudf column is type-erased with the type of the underlying data managed by its data type field rather than typing the memory pointers. This results in some confusion for type checkers (mypy) when a value is extracted from a column (with `to_py()`, for example) since as far as mypy knows that result could be of any type a column supports. We should see if there is a way to give mypy better hints for this based on the dtype of the Column. We do not want to add inheritance to pylibcudf columns just for this, so it may be the case that we cannot do anything about this aside from adding type ignores.