os-image-info takes too long to run (regression)

The evolution made in !6005 (merged) results in making os-images-info unit much slower to reconcile, because it now fetches metadata for a much larger number of OS image OCI artifacts.

But, in fact, there is only one case where it needs to fetch OS image OCI artifacts for older versions of Sylva: the use case where (a) we're in mgmt cluster context and (b) where the os-image-server unit is enabled.

We can easily improve the general cases (workload cluster context, or mgmt cluster context without os-image-server unit being enabled), by not having os-images-info unit process OS image OCI artifacts for older release of sylva unless the os-image-server unit is enabled.

/cc @rletrocquer