Bug in DataFrameMapper.get_names: unicode feature names

[AssafBenDavid]

Hi,

When the features returned by a transformer (by _get_feature_names) are unicode strings containing non-ascii characters, get_names throws a UnicodeEncodeError. This is due to line 256 in data frame_mapper.py:
return [name + '_' + str(o) for o in names]
This assumes any object in names can be converted to string using str.
It would be much better to use something like '%s_%s' % (name, o), which returns a unicode when needed. At the end, these strings become DataFrame column names, and DataFrames don't care about unicode indices.

[dukebody]

Thank you very much for the report and the contribution! Fixed in #165.