BUG: np.pad fails for sizes that do not fit in np.int_ #12023
Description
>>> np.pad(np.zeros((0, 0)), [(0, 2**30), (0, 0)], 'constant')
array([], shape=(1073741824, 0), dtype=float64)
>>> np.pad(np.zeros((0, 0)), [(0, 2**31), (0, 0)], 'constant')
ValueError: [(0, 2147483648), (0, 0)] cannot contain negative values.
>>> np.pad(np.zeros((0, 0)), [(0, 2**32), (0, 0)], 'constant')
array([], shape=(0, 0), dtype=float64)Switching an internal cast to intp solves the issue