Video streaming analytics

Sample code (this code sample is not ksql-ready, it needs to be adapted):

```
select hour, count(distinct(user_id)) as viewers
from video_activity
group by hour
order by hour
```

```
select hour, video_id, count(*) as views
from user_activity
where type = 'play'
group by video_id, hour
order by hour
```

```
select video_id, type, count(*)
from user_activity
where type = 'stop' or type = 'pause'
group by video_id, type
```

Idea: Maybe flag the videos that aren't being watched to the end?

Inspiration: https://databricks.com/notebooks/QoS/index.html#03.operational_dashboard.html