BUG: index attributes don't respect the width & need wrapping

[jreback]

I could get this only to show up in the notebook, but the idea is that the attributes (e.g. _format_attrs())
should be wrapped like we do data so that they don't get cut off.

In [14]: pd.options.display.width=20

In [15]: pd.CategoricalIndex(['foo','bar','baz','a really long string']*100,name='a long name')
Out[15]: 
CategoricalIndex([u'foo',
                  u'bar',
                  u'baz',
                  u'a really long string',
                  u'foo',
                  u'bar',
                  u'baz',
                  u'a really long string',
                  u'foo',
                  u'bar',
                  ...
                  u'baz',
                  u'a really long string',
                  u'foo',
                  u'bar',
                  u'baz',
                  u'a really long string',
                  u'foo',
                  u'bar',
                  u'baz',
                  u'a really long string'],
                 categories=[u'a really long string', u'bar', u'baz', u'foo'], ordered=False, name=u'a long name', dtype='category', length=400)