Fix DetachedInstanceError after StatusCodeRedirect

In certain cases, e.g. after an exception has been thrown, the database
session may be detached, but database objects managed by the orm are
still refered to from the same thread.

Such objects can be merged with the new db session, as described in
http://stackoverflow.com/a/9705458/201743

This is done with `c.user` which is re-retrieved from repoze.who after a
StatusCodeRedirect.

This fixes what ade22d7 works around.

See also discussion on: https://bitbucket.org/phihag/hhu-
adhocracy/issue/168/error-handler-causes-is-not-bound-to-a

adhocracy/lib/base.py

@@ -32,11 +32,12 @@ def __call__(self, environ, start_response):
         else:
             c.active_global_nav = 'home'
         c.user = environ.get('repoze.who.identity', {}).get('user')
-        try:
-            if c.user and (c.user.banned or c.user.delete_time):
-                c.user = None
-        except DetachedInstanceError, e:
-            log.exception(e)
+
+        # make sure we're not using a detached user object
+        c.user = model.meta.Session.merge(c.user)
+
+        if c.user and (c.user.banned or c.user.delete_time):
+            c.user = None
         c.active_controller = request.environ.get('pylons.routes_dict')\
             .get('controller')
         c.debug = asbool(config.get('debug'))