mgr/cephadm: Scale down ingress service only when the backend NFS service is scaled...

mgr/cephadm: Scale down ingress service only when the backend NFS service is scaled down

When scaling ingress down with a ranked NFS backend that still has more
daemons than its placement target, remove ingress from hosts without NFS
first and skip removing co-located ingress on NFS hosts until the next
iteration after NFS scales down.

Fixes: https://tracker.ceph.com/issues/76304
Signed-off-by: Shweta Bhosale <Shweta.Bhosale1@ibm.com>