Assignment to uint series drops datatype

Assigning a uint value to a slice of a uint series results in a dataframe of type int.

```python
s = pd.Series([1,2,3], dtype='uint64') # Also happens with other uint dtypes
s.loc[0] = 4
s
```
With Pandas 0.19.2, this returns:
```
0    4
1    2
2    3
dtype: uint64
```

With version 0.21.0, the dtype gets lost and replaced with int64:
```
0    4
1    2
2    3
dtype: int64
```