Pandas raise TypeError when function returns object-like results

This roll_apply raises TypeError: must be real number, not str when my customized function returns str / list or other object-like results. Because here it is: ndarray[float64_t] output

https://github.com/pandas-dev/pandas/blob/0691c5cf90477d3503834d983f69350f250a6ff7/pandas/_libs/window/aggregations.pyx#L1396