SIGTERM ignored - we should handle it

[consideRatio]

If this can run in less than 10 seconds, this issue is resolved.

docker build -t chp:latest . && \
echo "Starting..." && CONTAINER_ID=$(docker run -d chp:latest) && \
echo "Giving it time to startup..." && sleep 5 && \
echo "Stopping! Does this take less than 10 seconds?" && \
docker stop -t 10 $CONTAINER_ID

The reason this is an issue I care about, is because in k8s it takes 30 seconds we wait 30 seconds and then send SIGKILL to the first process. So if we make an upgrade to the proxy pod, we need to wait 30 seconds typically for it to actually terminate. We can make this be 0 seconds, but then we don't exit very nicely, so perhaps we could instead do like we do for SIGINT?

configurable-http-proxy/bin/configurable-http-proxy

Lines 367 to 372 in 5f912bb

	// trigger normal exit on sigint
	// without this, PID cleanup won't fire on SIGINT
	process.on("SIGINT", function() {
	log.warn("Interrupted");
	process.exit(2);
	});

This issue relates to jupyterhub/zero-to-jupyterhub-k8s#1476.

I'll submit a PR!