non-ascii bytes in "str" type are ambiguous.

Since the codepoints resulting from a decode depend on the encoding,
this test can only be an example of error handling, but the current code
doesn't treat it as as such. so IMHO this is wrong,  it breaks
another fix which I believe does the right thing in this respect,
for another issue. and so It is removed by this patch.
Unicode<-> bytes conversion and coercing should be done at I/O
entry points, not internal code.

Author: y-p

pandas/tests/test_format.py

@@ -246,7 +246,7 @@ def test_repr_corner(self):
 
     def test_frame_info_encoding(self):
         index = ['\'Til There Was You (1997)',
-                 '\xc1 k\xf6ldum klaka (Cold Fever) (1994)']
+                 'ldum klaka (Cold Fever) (1994)']
         fmt.set_printoptions(max_rows=1)
         df = DataFrame(columns=['a', 'b', 'c'], index=index)
         repr(df)