ENH: Support scalar operations against python `datetime` and `timedelta`

Proposed new feature or change:

import datetime as dt
import numpy as np

np.timedelta64(1, "D") + dt.timedelta(days=1)  # ✅ supported

array = np.array([np.timedelta64(1, "D"), np.timedelta64(2, "D")])
array + dt.timedelta(days=1)  # ❌ not supported
# ufunc 'add' cannot use operands with types dtype('<m8[D]') and dtype('O')

np.timedelta64 and np.datetime64 already support scalar operation against with the built-in datetime.datetime and datetime/timedelta types, however ndarray does not.

Reproduce the scalar and array examples from the issue and trace the ndarray ufunc handling behind the reported add error. Identify the relevant NumPy tests for datetime64/timedelta64 operations, then verify that the array expression works with Python datetime and timedelta values while existing scalar behavior remains unchanged.