multiindex xs slicing bug fix

[cpcloud]

#2903 solves the case for the first level but using the same variables from that issue, the following code raises a TypeError for the second level:

import pandas as pd
from numpy.random import rand
columns = pd.MultiIndex.from_tuples([('a', 'foo'), ('a', 'bar'), ('b', 'hello'), ('b', 'world')], names=['lvl0', 'lvl1'])
df = pd.DataFrame(rand(4, 4), columns=columns)
df.xs('foo', level='lvl1', axis=1)

This pull request fixes that.

[jreback]

looks ok to me

[ghost]

Added a vbench for xs, looks fine.