Backport: Raise AnsibleConnectionError on winrm connnection errors (#52226)

* Raise AnsibleConnectionError on winrm con errors

Currently all uncaught exceptions of the requests library that is used
in winrm will lead to an "Unexpected failure during module execution".

Instead of letting all exceptions bubble up we catch the connection
related errors (inkl. timeouts) and re-raise them as
AnsibleConnectionError so Ansible will mark the host as unreachable and
exit with the correct return code.

This is especially important for Zuul (https://zuul-ci.org) to
distinguish between failures and connection/host related errors.

* Update lib/ansible/plugins/connection/winrm.py

Co-Authored-By: westphahl <westphahl@gmail.com>

* Add changelog fragment

changelogs/fragments/winrm-ansible-conn-error.yaml

@@ -0,0 +1,3 @@
+---
+minor_changes:
+  - Raise AnsibleConnectionError on winrm connnection errors

lib/ansible/plugins/connection/winrm.py

@@ -134,6 +134,7 @@ try:
     import winrm
     from winrm import Response
     from winrm.protocol import Protocol
+    import requests.exceptions
     HAS_WINRM = True
 except ImportError as e:
     HAS_WINRM = False
@@ -479,6 +480,8 @@ class Connection(ConnectionBase):
                 raise AnsibleError('winrm send_input failed; \nstdout: %s\nstderr %s' % (to_native(response.std_out), to_native(stderr)))
 
             return response
+        except requests.exceptions.ConnectionError as exc:
+            raise AnsibleConnectionFailure('winrm connection error: %s' % to_native(exc))
         finally:
             if command_id:
                 self.protocol.cleanup_command(self.shell_id, command_id)