feat(router): optionally add jitter to config polling interval

[endigma]

Motivation and Context

Users with many server instances may experience burst load issues when all instances refresh their config simultaneously. This feature helps to mitigate this issue by adding a config option: poll_jitter that introduces variance on top of the base pollInterval. A value of 0 will result in no jitter, beyond inherent OS timing jitter.

Summary

• Adds a new function randomDuration to the controlplane package
• This is used within controlplane's Poll implementation to add variable jitter to the handler start time
• Tests assert the functionality of randomDuration and add some coverage to Poll as well
• A new config option poll_jitter has been created that represents some duration between 0 and its value to be added to the ticker's interval
• It defaults to 5s and cannot be less than 0s, a value of 0s will result in no jitter

Caveats of jitter implementation

The jitter is added between the tick being consumed and the handler being executed, which means the period the handler has to execute before a tick is dropped is not consistent. This is a non-issue where ticks being dropped is tolerable, but reusing the Poll implementation with jitter enabled and small intervals may surface this issue.

Checklist

• I have discussed my proposed changes in an issue and have received approval to proceed.
• I have followed the coding standards of the project.
• Tests or benchmarks have been added or updated.
• Documentation has been updated on https://github.com/wundergraph/cosmo-docs.
• I have read the Contributors Guide.

[endigma]

Should jitter now be a mandatory value with a non-zero default and enforced non-zero value?

[StarpTech]

I also thought about it. I think it is a good default to enable it. We have a default poll interval of 10s. We can have a jitter of 5s. You can validate min by adding a validation rule to the config json schema file.

[endigma]

rebased instead

[StarpTech]

Please update the PR description according to you latest changes.

[StarpTech]

The reason why the CI is failing is because the action is not allowed to run for forks. We would need to improve this. I run the tests locally and the PR is ready to merge to me. Thank you.