Large data values can raise value errors

When data values are too large to fit a tree-based model, the AutoML search can raise a value error. This issue is an edge case, so not a high priority but helpful to track.

```python
from numpy import array
from sklearn.ensemble import ExtraTreesRegressor

x = array([[ 7.91300786e+25],
[-5.75321932e+11],
[-1.67246806e+71],
[-7.55947560e+30],
[-2.93739824e+46]])

y = array([337000., 128000., 112000., 234000., 221000.])
ExtraTreesRegressor().fit(x, y)
```

```
ValueError: Input contains NaN, infinity or a value too large for dtype('float32').
```