Check for successful haproxy startup

[PKizzle]

Instead of assuming that haproxy API server is already reachable when starting up kubernetes-ingress wait until it is available. This fixes startup issues for systems running on low-end hardware where kubernetes-ingress would go into a crash-loop as haproxy never starts up fast enough.

[oktalz]

hi @PKizzle ,

thx for the PR.
if this is the issue on low-end hardware I agree we should address it.
However, we can't remove inotify from aux file, that file is allowed to be changed over time

[PKizzle]

Could you explain what you mean with the removal of inotify from the aux file? I am unable to find it in my code changes 😅

Edit: Are you maybe referring to my other PR?
Edit 2: Now I understand. It is regarding the commit 66b9efd. I have added the fsnotify based monitoring myself. As there was already an alternative change detection system in place I have removed the feature again.
Edit 3: I have removed the confusing commit history and merged it into one

[dkorunic]

This is much easier to fix with relying on S6 dependancies in S6 overlay 3.1 and with fine-tuning startupProbe. failureThreshold if needed. The changes here and in #593 bring too much changes for a specific case. I'll close this request and address this in incoming S6 overlay 3.1 overhaul.