fix: add auto_reconnect to the Postgrex.Notifications

[abc3]

If Postgres is not available at the time of startup, Postgrex.Notifications.start_link will kill the strategy process each time the supervisor restarts it, until all retries are exhausted.

libcluster_postgres/lib/strategy.ex

Line 60 in 82e36ab

{:ok, conn_notif} <- Postgrex.Notifications.start_link(state.meta.opts.()),

This PR fixes this behavior by starting Postgrex.Notifications with the :auto_reconnect option. This prevents the process from killing and allows it to keep retrying connection attempts in the background. Once Postgres becomes available again, the process will cast the node name into the cluster.

[abc3]

cc @Ziinc @filipecabaco

[abc3]

ah, auto_reconnect is already been added in #20