tests: fix long shutdown times for the webhook_receiver container (#5458)

Currently, `server.py` runs as PID 1, which means that it won't be
terminated by a `SIGTERM` signal unless it explicitly handles it (which
it doesn't). So when Docker tries to shut the container down, it sends
the server a `SIGTERM`, which gets ignored, and then sits there for 10
seconds before sending it a `SIGKILL`.

To work around this, enable the built-in Docker init program, which
forwards signals to the Python server. Since the Python server is no
longer PID 1, `SIGTERM` will now shut it down immediately.

The `init` option is supported starting from the Compose format version
3.7, so bump the version.

tests/docker-compose.test_servers.yml

@@ -1,10 +1,11 @@
-version: '3.3'
+version: '3.7'
 
 services:
   webhook_receiver:
     image: python:3.9-slim
     restart: always
     command: python3 /tmp/server.py
+    init: true
     env_file:
       - ./tests/python/webhook_receiver/.env
     expose: