cpg: fix issue when nodes with low id hang

cpg: fix issue when nodes with low id hang

The cpg config change callback where not made correctly for nodes
with a low member id (lowest id == master node) after corosync on
said node hung, due to IO, artifical suspend or other scheduling
related hangs.

This is releated to an heuristic for choosing/syncing the CPG member
downlist added in 6bbbfcb6b4af72cf35ab9fdb4412fa6c6bdacc12 (corosync
repository).

See whole issue thread:
https://lists.clusterlabs.org/pipermail/users/2018-March/014594.html

Upstream pull-request:
https://github.com/corosync/corosync/pull/347