Connection is not reestablished after ConnectionDropped

[srwilson]

I encountered this traceback running version 0.7:

Traceback (most recent call last):
  File "/opt/python/lib/python2.7/site-packages/kazoo/protocol/connection.py", line 551, in _send_request
    self._submit(request, connect_timeout, self._xid)
  File "/opt/python/lib/python2.7/site-packages/kazoo/protocol/connection.py", line 223, in _submit
    self._write(int_struct.pack(len(b)) + b, timeout)
  File "/opt/python/lib/python2.7/site-packages/kazoo/protocol/connection.py", line 236, in _write
    sent += bytes_sent
  File "/opt/python/lib/python2.7/contextlib.py", line 35, in __exit__
    self.gen.throw(type, value, traceback)
  File "/opt/python/lib/python2.7/site-packages/kazoo/protocol/connection.py", line 66, in socket_error_handling
    errno.errorcode[e.args[0]])
ConnectionDropped: ('socket connection error: %s', 'EPIPE')

All subsequent zookeeper calls raised ConnectionClosedError.

It seems the the client should attempt to reconnect after a transient connection loss such as this.

[cgordon]

We've been seeing this issue as well, but it's very rare. Here is my attempt to fix it, which we are testing now:

#41

[hannosch]

I fixed this using the patch in #41 - and added a test for the whole thing.