Rolling Primitives (mean, max, etc) do not work when cutoff_time is set

When a primitive like `RollingMax` or `RollingMean` is passed into DFS along with cutoff times, the resulting features do not correctly calculate the value of interest over the specified `time_frame`. Instead, the values show up in the feature matrix unchanged from their original value.

The following is an example of this:
```python
fm, f = ft.dfs(entityset=es, target_entity='temps',
trans_primitives=[RollingMax(time_frame='2d'),
RollingMean(time_frame='2d')],
return_variable_types='all',
max_depth=1,
cutoff_time=cutoff_times,
training_window='2d',
include_cutoff_time=True
)
fm.head()
```
The results look like this:

(Note this is the case for the `Entity` api or the featuretools 1.0.0 api with woodwork.

A simple fix appears to be adding `uses_full_entity = True` to each impacted primitive. We should confirm that this is, indeed, a fix and also look into why we didn't think `use_full_entity` was previously necessary.