describe() doesn't ignore Nan anymore

[rubennj]

.describe() is not ignoring the Nan values at percentiles, so it gives Nan:

http://pandas.pydata.org/pandas-docs/stable/basics.html#summarizing-data-describe

Before 0.18 it was working as expected:

http://pandas.pydata.org/pandas-docs/version/0.17.0/basics.html#summarizing-data-describe

[jreback]

dupe of #13098

fixed already by #13122 which will be in 0.18.2