fix(kuma-cp): don't leak goroutine on every tick in SimpleWatchdog

[lukidzi]

Checklist prior to review

While debugging some performance issue I've noticed that Kuma creates thousands of goroutines that might grow indefinitely. Whenever clients connect to KDS create an instance of SimpleWatchdog which is responsible for configuration generation. It's triggered every second and generates a snapshot which later puts into the cache. Lately, while implementing better context cancelation we introduced a change in a SimpleWatchdog which creates a new goroutine responsible for listing stop events and canceling the context. Goroutine is spawned in a loop and every second there was a new goroutine created which was listening events. In the case of KDS this code was running until the zone disconnects(disconnect cleaned up all the goroutines), and because connections are long-lasting it might create over 200k+ goroutines. It seems to be relevant to all the code using SimpleWatchdog. I am not sure if that fix handling all the cases correctly but while testing I've noticed that there is no more leak.

Before 1 zone connected for over a 1min:

After fix:

• Link to relevant issue as well as docs and UI issues --
• This will not break child repos: it doesn't hardcode values (.e.g "kumahq" as a image registry) and it will work on Windows, system specific functions like syscall.Mkfifo have equivalent implementation on the other OS --
• Tests (Unit test, E2E tests, manual test on universal and k8s) --
• Do you need to update UPGRADE.md? --
• Does it need to be backported according to the backporting policy? --
• Do you need to explicitly set a > Changelog: entry here or add a ci/ label to run fewer/more tests?

[github-actions]

backporting to release-2.3 with action

backporting to release-2.0 with action
backporting to release-2.2 with action