All and Any primitives propagate nulls

#### Bug/Feature Request Description
If you use `All` or `Any` aggregation primitive on a boolean feature which doesn't have any data to aggregate in a particular group then the resulting feature will have a null value after aggregation. This feature will also have an `object` dtype instead of `bool` or `float` if such action is intended. It may also result in errors while trying to apply such primitives twice but I haven't checked it.

I would prefer either missing values to be imputed with `False` by default or an option to set propagation to True/False/nan inside primitive function similar to setting unit in Time primitives.
#### Output of ``featuretools.show_info()``

```
Featuretools version: 0.9.1
Featuretools installation directory: C:\Users\user\AppData\Local\Continuum\
miniconda3\envs\cyclone\lib\site-packages\featuretools

SYSTEM INFO
-----------
python: 3.7.3.final.0
python-bits: 64
OS: Windows
OS-release: 2008ServerR2
machine: AMD64
processor: Intel64 Family 6 Model 45 Stepping 2, GenuineIntel
byteorder: little
LC_ALL: None
LANG: None
LOCALE: None.None

INSTALLED VERSIONS
------------------
numpy: 1.16.4
pandas: 0.23.4
tqdm: 4.32.2
toolz: 0.9.0
PyYAML: 5.1.1
cloudpickle: 1.2.1
future: 0.17.1
dask: 2.1.0
distributed: 2.1.0
psutil: 5.6.3
Click: 7.0
scikit-learn: 0.21.2
pip: 19.1.1
setuptools: 41.0.1
```