Better children tracking

[jabdoa2]

• Properly track all children
• Do not overwrite children if they are referenced multiple times
• Track which keys are required and which are optional in a configmap/secret
• Validate if our children have all required key (before: we only checked that the child existed)
• Add logging + kubernetes events for scheduling enable/disable
• Fix potential race in watcher code. We previously released the lock during watcher updates. There is a small chance that we might miss an updated child in a concurrent watcher. Not a huge chance but not impossible.

Fix #170
Fix #171
Fix #172

This fixes a few cases we found in production where pods were restarting after creation (in combination with cert-manager secrets which initiall miss a key). Other than that it performed very well.

On top of #169. We might want to merge that first.

[toelke]

As an aside: You know that github handles stacked PRs rather well? If you change the base of this PR to a feature branch, it will automatically change to go on top of main when that is merged

[jabdoa2]

As an aside: You know that github handles stacked PRs rather well? If you change the base of this PR to a feature branch, it will automatically change to go on top of main when that is merged

I did not know. Unfortunately, I cannot select my other PR since the branch technically is in my github org and this PR is against the wave-k8s org. Should that still work?

[toelke]

Sorry, yes, that makes sense 🤦‍♂️