Allow specifying normalized form for Unicode filter

When initializing the `Unicode` filter, allow the developer to specify which normalized form to use.  E.g.:

```python
In [1]: f.Unicode().apply('Ame\u0301lie')
Out[1]: 'Am\xe9lie'

In [2]: f.Unicode(normalize='NFD').apply('Ame\u0301lie')
Out[2]: 'Ame\u0301lie'
```