e2e: Improve AWS cleanup job

Looking at our AWS environment, there are some Elastic IPs which have no association but have not been released. They all have the `sigs.k8s.io/cluster-api-provider-aws/role: common` tag, which indicates they come from CAPA testing. It's unclear why this happens, as there are no records we can use at the moment. However it would be best if the [cleanup job](https://github.com/rancher/turtles/blob/main/.github/workflows/e2e-cleanup.yaml) also takes Elastic IPs in account, trying to delete them. It would also be best to spot when this happens for further investigation, maybe using `continue-on-error: true` and let this action step fail if an Elastic IP is found, before deleting it. This allows the user to collect the run logs and try to verify why CAPA did not delete the resource.