BUG: fillna returns frame when inplace=True if fill values are specified as a dict

`fillna` will return the frame in question if passed fill values via a dict, even if `inplace=True`.

```
In [6]: data = pd.DataFrame({'a': [1, None], 'b': [2, None]})

In [7]: data.fillna({'a': 0}, inplace=True)
Out[7]: 
     a    b
0  1.0  2.0
1  0.0  NaN

In [8]: data
Out[8]: 
     a    b
0  1.0  2.0
1  0.0  NaN

```