clean up calico dhcp state on reboots (backport #952)

[mergify]

Problem:

Under certain scenarios when a harvester node is hard rebooted the canal cni can leak ip addresses
https://docs.rke2.io/known_issues#canal-and-ip-exhaustion
Over time these may build up and result in ip address pool exhaustion for a node since.

The only option in such a scenario is to clean up un-used ip leases from /var/lib/cni/networks/k8s-pod-network

Solution:

PR introduces a minor change to add directives for cni reset in base image.
This ensures they are rolled out to existing systems as part of the upgrade cycle

Related Issue:
harvester/harvester#7471

Test plan:

• Install harvester from current changes to a node
• Once harvester is running hard reboot the node
• Once harvester is up and running check the timestamp on the folder

/system/oem # ls -ld /var/lib/cni/networks/k8s-pod-network
drwxr-xr-x 2 root root 4096 Feb  7 02:40 /var/lib/cni/networks/k8s-pod-network

• the folder would have been recreated after the reboot

/system/oem # last reboot
reboot   system boot  5.14.21-150500.5 Fri Feb  7 02:32   still running

wtmp begins Fri Feb  7 02:32:03 2025
```<hr>This is an automatic backport of pull request #952 done by [Mergify](https://mergify.com).