We were sending lastTransaction requests while register requests were
in-flight.  If register failed, then the lastTransaction request was
invalid, causing the connection to be closed. :(

When we update the server, we'll have register return lastTransaction
and probably info, since the client wants that information on connect.

To make matters worse, when the other site closes a connection, None
is passed to protocol connection_lost methods, which was fouling up
sornstream error detection.

Because it's only accessed by the asyncio thread.

Don't test obsolete method ``invalidate``. It's not used anymore in
master and not present on this (asyncio) branch.

Removed the reuable tests because transaction buffers are no-longer reused.

- testZEO tests now pass

- async tests now pass again

  Probably need to write more async tests to reflect changes.
  (Or maybe the ZEO tests that drove tem are enough.)

- dropped heartbeat tests, which were insane. Will add simpler test
  when I add heartbeats to the async implementation.

Many tests passing. Quite a few still failing.

- Issue with notify_connected, ClientStorage wants to make requests in
  response to being notified.  This is problematic because
  synchronsouse calls cause deadlock in this situation as do
  asyncronous calls done in a multi-threaded fashion.

  - Call get_info from io thread during startup, because
    notify_connected wants it.

  - Added an same-thread asyncronous API.

  - Added comment warning of this issue.

  - Added a little more logging.

- fixed an ordering issue when protocol is disconnected. It should
  notify the client before it cleans up it's futures to prevent
  getting more.

- Expose protocol_version to client so it can adjust it's behavior to
  the .

- More logging

- Fixed tpc_finish:

  - Use tid from server to update cache.

  - Accept and call callback function.

- Implemented flow control

- Added connection/disconnection notification (to client storage).

- implemented get_peername.

- implemented is_read_only

- renamed callAsync to async (death to Camels!)

refinements (including spelling and considering allowing tpc_finish and loads to be in flight at once.

Renamed connection_timeout to connect_poll and use it when reconnecting.

Optimized hanfline of first messagem, as we did in zrpc, because it
only occurs once. :)

More/better comments.

Move loop argument to front of constructor arguments.

Added close/close_threadsafe to wait for connection on close.

Added is_connected.

Added new_addr.

Also got rid of the adapter machinery. It didn't buy enough to justify
the wrapping.

- Moved cache into async thread to avoid lots of locking.

- Setup delegation to storage.

- Provide thread wrapper that runs the async protocol in a thread.

Much left to do, but this is a nice little spike.

Add support for PyPy.

Let checkQuickVerificationWith2Clients use a longer timeaut and raise an assertion error if the timeout fails. Trying to debug intermittant failures on Travis, seen in all versions of Python.

Per @mgedmin, use a renormalizing pattern to fix the unicode string matchin in Python3. Also just print the data for the storage we expect, because the missing u prefixes change the spacing when pprint nests dicts and things don't match under Python2.

Fix the ruok protocol under Python3: a bytes/unicode problem. However, the doctest cases still don't pass because they have leading u prefixes in them.

One more place to account for the socket.error switchero on Python3. Still doesn't fix nagios/ruok though.

Make the nagios metrics come back in a deterministic order. This fixes the nagios test under PyPy, but they still fail under Python 3 for the same reason test_ruok does.

test_ruok was hanging under Python3. Now it doesn't hang but just fails with 'struct.error: unpack requires a bytes object of length 4'

Fix some test failures under Python3.3+. This loop can also raise a RuntimeError([Errno 9] Bad file descriptor), though, so not all are fixed. But this may be enough to fix the hang.