CI: Extend timeouts for RAX

It appears our jobs end up encountering some performance issues
on some RAX hosts where the overall runtime is already close to
the maximum for the job window.

Ultimately what happens is the performance ends up putting the boot
of the workload VM just a little bit later and we end up failing
the job as the VM is still early in booting up.

Change-Id: I309220d7f61f87b7457668fcf586d4ea152d363b

.zuul.yaml

@@ -123,7 +123,7 @@
         Integration job using Glance as image source and CentOS8 with local boot
         and UEFI. Images are built with diskimage-builder.
     parent: metalsmith-integration-base
-    timeout: 8200
+    timeout: 8500
     run: playbooks/integration/centos8-integration.yaml
     vars:
       devstack_localrc:

playbooks/integration/exercise.yaml

@@ -79,7 +79,7 @@
   register: ssh_result
   until: ssh_result is success
   retries: 30
-  delay: 20
+  delay: 30
 
 - name: Undeploy a node
   command: metalsmith --debug undeploy --wait 900 test