Apply Across Non-Unique Indices.

Applying a function across a data-frame does not seem to work for non-unique indices.

```
df = pandas.DataFrame([[1,1,1], [2,2,2], [3,3,3]], index=['a','a','c'])
df.apply(lambda s: s[0], axis=1)

a    2
a    2
c    3
```

It seems to be just returning the value for the first instance of the index.
