ER: Uninformative error message when concat-ing Series with frames

```
In [164]: df = DataFrame(randn(10, 1))

In [165]: concat([df[0], df], axis=1)
```

throws an `AttributeError`. It should just say something about incompatible types in `concat`.

I wonder if it's worth allowing mixed types where it makes sense. E.g., `concat`-ing a `Series` and a `DataFrame`. It might be more consistent with the fact that `Series` is now an `NDFrame`.
