[CFT] math/py-pandas: Update to 0.11.0rc1

[John W. O'Brien]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Thanks to y-p over at the pandas github project, I have isolated the
root cause of both test failures. While I will continue to monitor the
upcoming release, and contribute as it seems necessary, final
resolution is now in upstream hands.

This is a summary for the benefit of freebsd-python at .

1. pandas.tests.test_format
   TestDataFrameFormatting.test_to_string_repr_unicode

The reason this test was failing due to some long-standing bad
behavior in py-gtk that affects pandas via matplotlib when a GTK
backend is in use. pandas uses sys.getdefaultencoding() to initialize
its own 'display.encoding' option, which affects behaviors throughout
pandas, and (in this case) in the pandas.core.format module. On
import, the gtk module changes the default encoding, which should
otherwise remain constant after python processes site.py.

If pandas is imported and then gtk is imported, any code that depends
on the value of the 'display.encoding' option may break. An
operational work-around is to import gtk before importing pandas.

Detecting and accommodating the breakage caused by py-gtk appears to
be non-trivial, so the pandas devs would be right to elect not to
implement it for the impending 0.11.0 release.

2. pandas.tests.test_index
    TestMultiIndex.test_legacy_pickle

The problem here is that the files in the 0.11.0rc1 tarball have CRLF
newlines, including a sample pickle file this test uses. There are no
CRLFs in the main repo.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)

iQEcBAEBAgAGBQJRc2DmAAoJEBRzAKlhyP/FGWIIANrP+nIuumnKkpTElg3oEvJu
B3LIBuJvVPiaFMssipu/SU4ZaHx1zKZL6gyY5Q/UfX8YIu1tfFQIMoZfUm8xx8iH
+TyL3HacK3TauJnwtj3S7kIWfHM30+gmgMsbphnb9pngpXkRKfjSx0sk0WW98GA/
6THgmCZpq5YGZWKyzMzaNSlTNGVvrGqnGdiJksVrIvQx8SB1/Js53ofiA/6Iype8
+ql//Mz5h2ysJ5GQiaFYyzUh5EhbtTy8VtaN2JXr6g9Ua6bqxXHjaJVHxOPIZPQ4
4CiaH8qD7GCuHdMQ8EREp9zEaRTgIVQoKGFGNeMhZIMKChjeAAwUPlI5kBtNnvs=
=wnkP
-----END PGP SIGNATURE-----