BUG: int sparse series is converted to float if sliced

This looks weird:

``` python
In [1]: pd.__version__
Out[1]: '0.14.1-388-g54678dd'

In [2]: sp_series = pd.Series([0,0,1]).to_sparse(fill_value=0)

In [3]: sp_series.dtype
Out[3]: dtype('int64')

In [4]: sp_series[::-1].dtype
Out[4]: dtype('float64')

In [5]: sp_series[::2].dtype
Out[5]: dtype('float64')

In [6]: sp_series[:-1].dtype
Out[6]: dtype('float64')
```
