Column names as class attributes only work if type hint is wrapped in `Series`

[GiorgioBalestrieri]

This relates to #364 (which introduced a great feature).

When accessing column names as class attributes of a schema defined by inheriting from DataFrameModel, if the type hint is not wrapped in Series, then the type of the column name is inferred to be the column type, instead of str.

Minimum working example:

import pandera as pa
from pandera.typing import Series


class FooSchema(pa.DataFrameModel):
    a: Series[int]
    b: Series[float]


class BarSchema(pa.DataFrameModel):
    x: int
    y: float


if __name__ == "__main__":
    foo_columns: list[str] = [FooSchema.a, FooSchema.b]
    bar_columns: list[str] = [BarSchema.x, BarSchema.y]

Mypy output:

pandera_issue.py:17: error: List item 0 has incompatible type "int"; expected "str"  [list-item]
pandera_issue.py:17: error: List item 1 has incompatible type "float"; expected "str"  [list-item]
Found 2 errors in 1 file (checked 1 source file)

This is relatively easy to solve by wrapping the type hints in Series, but:

1. I would say this is unexpected (and undocumented) behavior, and those error messages are quite confusing
2. having to wrap every type hint in Series adds boilerplate (which is why support for non-wrapped type hints was added in the first place)

EDIT: I think the reason Series[...] works is that SeriesBase (that Series inherits from) implements __get__ with str as a return type:

pandera/pandera/typing/common.py

Lines 159 to 161 in 660095b

	def __get__(
	self, instance: object, owner: type
	) -> str: # pragma: no cover