client: fix simultaneous (re)connections to the master

This fixes a reqression in commit c39d5c67,
which could leads to failures like:

2016-11-29 09:56:58,756 ERROR ZODB.Connection Couldn't load state for 0x4843
Traceback (most recent call last):
  File "ZODB/Connection.py", line 860, in setstate
    self._setstate(obj)
  File "ZODB/Connection.py", line 901, in _setstate
    p, serial = self._storage.load(obj._p_oid, '')
  File "neo/client/Storage.py", line 82, in load
    return self.app.load(oid)[:2]
  File "neo/client/app.py", line 352, in load
    data, tid, next_tid, _ = self._loadFromStorage(oid, tid, before_tid)
  File "neo/client/app.py", line 372, in _loadFromStorage
    for node, conn in self.cp.iterateForObject(oid, readable=True):
  File "neo/client/pool.py", line 91, in iterateForObject
    pt = self.app.pt
  File "neo/client/app.py", line 146, in __getattr__
    return self.__getattribute__(attr)
AttributeError: 'Application' object has no attribute 'pt'

neo/client/app.py

@@ -138,11 +138,11 @@ class Application(ThreadedApplication):
 
     def __getattr__(self, attr):
         if attr in ('last_tid', 'pt'):
-            if self._connecting_to_master_node.locked():
-                if attr == 'last_tid':
-                    return
-            else:
-                self._getMasterConnection()
+            self._getMasterConnection()
+            # XXX: There's still a risk that we get disconnected from the
+            #      master at this precise moment and for 'pt', we'd raise
+            #      AttributeError. Should we catch it and loop until it
+            #      succeeds?
         return self.__getattribute__(attr)
 
     def log(self):

neo/client/handlers/master.py

@@ -94,13 +94,14 @@ class PrimaryNotificationsHandler(MTEventHandler):
 
     def answerLastTransaction(self, conn, ltid):
         app = self.app
-        if app.last_tid != ltid:
+        app_last_tid = app.__dict__.get('last_tid', '')
+        if app_last_tid != ltid:
             # Either we're connecting or we already know the last tid
             # via invalidations.
             assert app.master_conn is None, app.master_conn
             app._cache_lock_acquire()
             try:
-                if app.last_tid < ltid:
+                if app_last_tid < ltid:
                     app._cache.clear_current()
                     # In the past, we tried not to invalidate the
                     # Connection caches entirely, using the list of