avoid node rolling update on a change of metallb settings
Today, since Flux manages MetalLB, we don't need to trigger a node rolling update when MetalLB configuration is changed.
However this does happen, because we inject MetalLB configuration into CAPI manifests for the code handling the creation of the first node, with the unneeded/unwanted side-effect of trigerring a node rolling update when MetalLB configuration changes.
This could be skipped to avoid a rolling update.
(related to similar issue with calico: #1981 (closed))
Edited by Thomas Morin