confusion_matrix doesn't validate normalize keyword

```python
import numpy as np
from sklearn.metrics import confusion_matrix
confusion_matrix(np.random.randint(0, 5, size=100),
                 np.random.randint(0, 5, size=100), normalize='ohnodontdoit')
```

should raise an error but runs fine with no normalization. I found this because I did ``normalize=True`` and it didn't do anything and didn't raise an error.