[main] Always reset failure counts on existing plans

[HarrisonWAffel]

Issue: #47652

Problem

In all previously released versions of Rancher, the rancher-wins service does have its underlying binary automatically upgraded to the latest version after an upgrade of the Rancher server. This behavior differs from Linux, and forces users to reprovision their Windows nodes in order to receive the latest enhancements within the system-agent (which is embedded in rancher-wins). This has resulted in unnecessary conditional logic within Rancher to account for behavior differences on nodes joined to the same cluster.

While implementing #46620 for Windows nodes, an issue was encountered where plan failures would not be accurately reported as out-dated versions of rancher-wins did not contain the necessary system-agent changes. To address this, Rancher was updated to conditionally update the failure-count property on Windows plans such that it is only set for newly provisioned nodes running the latest version of rancher-wins.

This change was effectively a stop gap, and should be considered as tech debt.

Solution

Revert the changes made in #47651. The changes raised in rancher/wins#260 will update the windows SUC image to automatically upgrade the underlying wins.exe binary used by rancher-wins, ensuring that the proper version of the system-agent is installed onto all Windows nodes.

Testing

• Provision a windows cluster using a commit associated released version of Rancher, like 2.9.3
• Ensure that the windows node plan failure count is set to 1
• Upgrade the rancher server to a commit running a newer version of rancher, like 2.9.4 or 2.10
• Ensure that the windows node plan failure count it updated to 5

Engineering Testing

Manual Testing

I've done the above

Automated Testing

QA Testing Considerations

Regressions Considerations