Installing a BareMetal IPI cluster using provisioningNetwork set to `Managed`, gets the cluster installed successfully and a route is created via the provisioning IP on the pod scheduling the metal3 pod. 

~~~
Working state; Route created - 

192.168.xx.0/20 dev --readacted-- proto kernel scope link src 192.168.xx.10

metal3-static-ip-set container - 

 initContainers:
    - resources:
      ..
      name: metal3-static-ip-set
      command:
        - /set-static-ip
      env:
        - name: PROVISIONING_IP
          value: 192.168.xx.10/20
        - name: PROVISIONING_INTERFACE
~~~

But, after switching from provisioningNetwork `Managed` to `Unmanaged`, there is no route created via the provisioning IP. Also the metal3-static-ip-set yaml has incorrect subnet set for provisioning network. The expected on is /20, but it gets /32.

~~~
No route created with unmanaged ProvisioningNetwork :

..
spec:
  ..
  provisioningIP: 192.168.xx.10
  provisioningNetwork: Unmanaged
  provisioningNetworkCIDR: 192.168.xx.0/20
  ..

  initContainers:
    - resources:
      ..
      name: metal3-static-ip-set
      command:
        - /set-static-ip
      env:
        - name: PROVISIONING_IP
          value: 192.168.xx.10
        - name: PROVISIONING_INTERFACE
          ..
~~~

Tried creating NNCP via kubernetes nmstate operator for each master node. The metal3's metal3-static-ip-set init container crashes with: 

~~~
+ '[' -z 192.168.xx.10 ']'
+ '[' -z vlanyyy ']'
+ '[' -z vlanyyy ']'
++ ip -o addr show dev vlanyyy scope global
+ '[' -n '15: vlanyyy inet 192.168.xx.11/20 brd 192.168.xx.xy scope global noprefixroute vlanyyy\ valid_lft forever preferred_lft forever' ']'
+ provisioning_net_valid
+ local prov_vip_prefixlen
++ ipcalc --no-decorate --prefix 192.168.xx.10
+ prov_vip_prefixlen=32
++ ip -j -o addr show dev vlanyyy
++ jq -r '.[0].addr_info[] | select(.scope == "global").local'
+ for ip_addr in $(ip -j -o addr show dev "${PROVISIONING_INTERFACE}" | jq -r '.[0].addr_info[] | select(.scope == "global").local')
+ local host_ip_cidr=192.168.xx.11/32
+ ipcalc --check 192.168.xx.11/32
++ ipcalc --no-decorate --network 192.168.xx.11/32
++ ipcalc --no-decorate --network 192.168.xx.10
+ '[' 192.168.xx.11 = 192.168.xx.10 ']'
+ echo 'ERROR: provisioning interface "vlanyyy" IP addresses are all outside the provisioning subnet "192.168.xx.10"'
ERROR: provisioning interface "vlanyyy" IP addresses are all outside the provisioning subnet "192.168.xx.10"
+ return 1
+ exit 1
~~~   

Install a BareMetal IPI cluster with ProvisioningNetwork set to Managed. Switch provisioning network to unmaaged and add NNCP for master nodes for provisioning interface. 

1. Install a BareMetal IPI cluster with ProvisioningNetwork set to Managed.    
2. Change the provisioningNetwork to unmanaged 
3. Add ip and routes via NNCP (kubernetes-nmstate operator) for all master nodes.   
4. Check the the metal3 pod's metal3-static-ip-set init container.     

The metal3 pod fails with errors. 

A route should be created with the NNCP in place. 

- Discussion in slack thread - https://redhat-internal.slack.com/archives/CFP6ST0A3/p1765874702761259