BUG: incorrect casting in ndarray + DataFrame

[jbrockmendel]

tdi = pd.timedelta_range('1s', periods=10)
df = pd.DataFrame({'a': tdi, 'b': tdi})
arr = np.arange(2)

>>> df + arr
[...]
AssertionError
# should be TypeError, but close enough for government work

>>> arr + df
                a                      b
0 0 days 00:00:01 0 days 00:00:01.000000
1 1 days 00:00:01 1 days 00:00:01.000000
2 2 days 00:00:01 2 days 00:00:01.000000
3 3 days 00:00:01 3 days 00:00:01.000000
4 4 days 00:00:01 4 days 00:00:01.000000
5 5 days 00:00:01 5 days 00:00:01.000000
6 6 days 00:00:01 6 days 00:00:01.000000
7 7 days 00:00:01 7 days 00:00:01.000000
8 8 days 00:00:01 8 days 00:00:01.000000
9 9 days 00:00:01 9 days 00:00:01.000000