engine: Fix started libvirt daemon requirement

This patch makes virt-manager usable again without systemd (or dbus).
When started without systemd, but with libvirtd properly working,
there was error reported, and virt-manager didn't even try to connect
to the destination.

This patch modifies the behavior in two ways.  When connection is
added with "do_start=False" it will still try to connect, but won't
error out when unsuccessful.  The second modification is that upon
startup, the idle_connect() function will error out only if we failed
starting libvirtd _and_ we couldn't connect to the uri.  So if we
failed starting libvirtd, but could connect, there is no disruption
for the user.

virtManager/engine.py

@@ -217,10 +217,10 @@ class vmmEngine(vmmGObject):
         # packagekit async dialog has a chance to go away
         def idle_connect():
             do_start = packageutils.start_libvirtd()
-            if not do_start:
+            connected = self.connect_to_uri(tryuri, autoconnect=True, do_start=do_start)
+            if not connected and do_start:
                 manager.err.ok(_("Libvirt service must be started"), warnmsg)
 
-            self.connect_to_uri(tryuri, autoconnect=True, do_start=do_start)
         self.idle_add(idle_connect)
 
 
@@ -489,6 +489,11 @@ class vmmEngine(vmmGObject):
 
             if do_start:
                 conn.open()
+            else:
+                try:
+                    conn.open()
+                except:
+                    return None
             return conn
         except Exception:
             logging.exception("Error connecting to %s", uri)