CI: Increase galera node timeouts

Some multinode jobs have been failing due to galera suspecting inactive nodes
Tweak it as per [1].

[1]: https://galeracluster.com/library/documentation/recovery.html

Change-Id: I5577ea2c23d6dbd440bd52899a30ea9531996256

tests/templates/globals-default.j2

@@ -50,9 +50,12 @@ docker_custom_config:
 distro_python_version: "{{ kolla_python_version }}"
 {% endif %}
 
-# MariaDB/Galera - fine tune gmcast.peer_timeout
+# MariaDB/Galera - fine tune timeouts
 mariadb_wsrep_extra_provider_options:
   - "gmcast.peer_timeout=PT15S"
+  - "evs.suspect_timeout=PT10S"
+  - "evs.inactive_timeout=PT30S"
+  - "evs.keepalive_period=PT3S"
 
 nova_compute_virt_type: "{{ virt_type }}"