GradientBoostingRegressor always sets presort to auto 

In the call to super inside `__init__()` `GradientBoostingRegressor` does not pass thru the presort variable to the superclass:

```
    super(GradientBoostingRegressor, self).__init__(
        loss=loss, learning_rate=learning_rate, n_estimators=n_estimators,
        ...
        max_leaf_nodes=max_leaf_nodes, warm_start=warm_start,
        presort='auto')
```
