Switch from updating to patching the target

[iroller]

Fix the race condition for target resources that can be changed during CPA operations:

Error while autoscaling: failed to scale resource: failed to apply scaling changes to resource:

Operation cannot be fulfilled on deployments.apps:
the object has been modified; please apply your changes to the latest version and try again.

One of the common reasons for this issue is status.availableReplicas field being updated multiple times a second in large churning Deployments. Each time resource changes Kubernetes updates metadata.resourceVersion and rejects the update if resource versions do not match.

Switching to patching only the replica count helps to avoid the issue altogether. This patch should work for all supported target resource types.

[jthomperoo]

This all looks good to me - thanks for the explanation of the change and the pull request, well written and easy for me to follow.

Merging now, will cut a new version v2.9.0 with these changes in it.

[jthomperoo]

Now available in v2.9.0 - thanks for submitting this change.