Fix information disclosure caused by unreachable nodes

Currently we can leak secrets if we encounter unreachable nodes
combined with a task using with_items and no_log. In this case the
item variables are written to both the job-output.json and
job-output.txt. Upstream Ansible has the same issue [1].

The text log can be fixed by defining the v2_runner_on_unreachable
callback the same as v2_runner_on_failed.

The json log can be fixed the same way as the upstream Ansible issue.

[1] https://bugzilla.redhat.com/show_bug.cgi?id=1588855

Change-Id: Ie5dd2a6b11e8e276da65fe470f364107f3dd07ef

zuul/ansible/callback/zuul_json.py

@@ -137,7 +137,7 @@ class CallbackModule(CallbackBase):
 
     def v2_runner_on_ok(self, result, **kwargs):
         host = result._host
-        if result._result.get('_ansible_no_log', False):
+        if result._result.get('_ansible_no_log', False) or result._task.no_log:
             self.results[-1]['tasks'][-1]['hosts'][host.name] = dict(
                 censored="the output has been hidden due to the fact that"
                          " 'no_log: true' was specified for this result")

zuul/ansible/callback/zuul_stream.py

@@ -606,3 +606,5 @@ class CallbackModule(default.CallbackModule):
                 delegated_host=delegated_vars['ansible_host'])
         else:
             return result._host.get_name()
+
+    v2_runner_on_unreachable = v2_runner_on_failed