Clarify uncertain nature of tracebacks on Python 2 (#42513)

* Only add exception/traceback on Python 3

On Python 2 the traceback could be any exception from the stack frame
and likely unrelated to the fail_json call.

On Python 3 the traceback is cleared outside any exception frame, so the
call always returns the most inner traceback (if any), and therefor is
most likely related to the fail_json call.

* Add uncertainty to traceback on Python 2

On Python 2 the last exception in the stack frame is being returned,
this could be unrelated to the actual error, especially if fail_json()
is called outside an except: block.

lib/ansible/module_utils/basic.py

@@ -2384,9 +2384,14 @@ class AnsibleModule(object):
             raise AssertionError("implementation error -- msg to explain the error is required")
         kwargs['failed'] = True
 
-        # add traceback if debug or high verbosity and it is missing
+        # Add traceback if debug or high verbosity and it is missing
-        # Note: badly named as exception, it is really always been 'traceback'
+        # NOTE: Badly named as exception, it really always has been a traceback
         if 'exception' not in kwargs and sys.exc_info()[2] and (self._debug or self._verbosity >= 3):
+            if PY2:
+                # On Python 2 this is the last (stack frame) exception and as such may be unrelated to the failure
+                kwargs['exception'] = 'WARNING: The below traceback may *not* be related to the actual failure.\n' +\
+                                      ''.join(traceback.format_tb(sys.exc_info()[2]))
+            else:
                 kwargs['exception'] = ''.join(traceback.format_tb(sys.exc_info()[2]))
 
         self.do_cleanup_files()