mgr/cephadm: Disable keepalived by default and start during serve loop if needed

mgr/cephadm: Disable keepalived by default and start during serve loop if needed

If the ingress service is configured with a placement count lower than the number of hosts,
then when one node goes down, Keepalived is configured on the remaining hosts.
When the failed host comes back up, if it was previously configured as the master,
the VIP may become active on two hosts. To prevent this, Keepalived is disabled by default
and started within the service loop.

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