[BUG] kapacitor alerts are not running as child processes

When using the `exec` handler of kapacitor, the script triggered by kapacitor is not running as a child process of `kapacitord`, which itself has pid1.

In my use case, I have scripts that handle a longer chain in pushover (send alert, wait for acknowledgement, send to another user...)

If the docker container is stopped within that runtime of the script, my script is silently and ruggedly killed - as there is no TERM signal I can handle with.

SOLUTION: All the alerts triggered by kapacitor should run as child processes of kapacitord or another process with pid1 handling all the subprocesses.

Showing the processlist with an alert running `sleep 30`, just to show the missing process tree:

```
root 1 4.9 0.1 2706372 161568 ? Ssl 17:43 0:44 kapacitord
root 19575 0.0 0.0 17976 2824 ? S 17:58 0:00 /bin/bash /usr/local/bin/pushover-alert-chain.sh
root 19576 0.0 0.0 4184 612 ? S 17:58 0:00 \_ sleep 30
```

Image:` kapacitor:1.5.4`

Start by examining Kapacitor's exec handler and how kapacitord launches alert scripts in the kapacitor:1.5.4 Docker image. Reproduce the process tree with a script that runs sleep 30, then observe shutdown behavior and signal delivery. Done means an exec alert remains a child process of the service and can handle TERM when the container stops.

Written by the indexing model from the issue text.