Fix poll jittering in Object controller

[aerfio]

Description of your changes

This PR fixes Object controller which sometimes returns a negative poll interval if the reconciled Object instance is not ready, which is then ignored by controller-runtime internals.

In my case this has been observed in a situation in which the k8s object from spec.forProvider.manifest takes a while to get ready, but under 10m which is the default pollInterval.

The logic which calculates the pollInterval does not take into account that the base pollInterval has been updated to 30 seconds if the reconciled resource is not ready. It took the base value of pollJitter which is 1 minute, and calculated next poll interval using this formula:

return pollInterval + time.Duration((rand.Float64()-0.5)*2*float64(pollJitter))

I guess that most users run provider-kubernetes without adjusting default poll interval, so if my math is right if rand.Float64 returns anything less then 0.25 (its documentations states: "// Float64 returns, as a float64, a pseudo-random number in the half-open interval [0.0,1.0)"), the returned value is less then 0. 25% chance to run into this bug. And if I understand the c-r source code right, it ignores negative RequeueAfter.

The solution is to recalculate the pollInterval taking into account the pollJitterPercentage - this means that with default value of poll jitter precentage the poll interval would be in half-open interval [27.0, 33.0) (seconds).

I have:

• Read and followed Crossplane's contribution process.
• Run make reviewable test to ensure this PR is ready for review.

How has this code been tested

[turkenh]

Great catch 🙌 and awesome PR description 🙇‍♂️ Thanks a lot!

[github-actions]

Successfully created backport PR #275 for release-0.14.

[aerfio]

@turkenh Hi, can I count on a patch release with this fix? 🙏🏻

[turkenh]

@aerfio sure, https://github.com/crossplane-contrib/provider-kubernetes/releases/tag/v0.14.1 is out.

[aerfio]

@turkenh Thank you!