fix: delete endpoint on cni state save failure

[QxBytes]

Reason for Change:

In windows, it is possible to successfully create an HNS endpoint and then subsequently fail to write to the CNI statefile. While the CNI call fails, the HNS endpoint is never deleted, causing retries to fail because the HNS endpoint has already been created. As it is not possible to retrieve the existing endpoint on a retried ADD call, we instead opt to delete the HNS endpoint if saving the CNI state fails.

Issue Fixed:

See above.

Requirements:

• uses conventional commit messages
• includes documentation
• adds unit tests
• relevant PR labels added

Notes:
The code change affects both the windows and linux scenarios (if saving fails, deleteEndpoint will call the platform specific version for deletion of the endpoint). On windows, tested triggering a save failure and then examining HNS endpoints-- the endpoint is cleaned up which is confirmed in the logs. Without the fix, the endpoint still shows and produces the endpoint already exists (ip address claimed) error. On linux transparent-vlan mode, tested triggering a save failure, which does call into deleting the endpoint. Subsequent delete/adds for that pod run without erroring.