This increases cluster on-line flexibility, allowing new masters to be
added without needing a restart of all masters.
Downside is that more responsibility is pushed to cluster admins: they must
update node configurations so they contain enough master nodes to allow
them to join cluster after a restart.

See http://bugs.python.org/issue13361 ("getLogger does not check its argument").

Custom ctypes wrapper for epoll is not needed with python 2.6+.
Also, do not retry upon temporary error: it is valid to handle no packet,
and retrying the poll call would increase overall timeout without easy fix.
Also, remove meaningless frozensets (leftovers from an old bug's symptoms).

Also simplify some code by using some new Python syntax.

This hask is mainly for performance tests, which should be run on a physical
machine that doesn't do anything else.

Commit 5c5c7d9a was a wrong fix because RLock
does not raise 'thread.error' like Lock (when releasing an unacquired lock).

Python >= 2.5 raises a RuntimeError and older Python raise an AssertionError.

Conflicts:
	CHANGES

- Change protocol to use SHA1 for all checksums:
  - Use SHA1 instead of CRC32 for data checksums.
  - Use SHA1 instead of MD5 for replication.

- Change DatabaseManager API so that backends can store raw data separately from
  object metadata:
  - When processing AskStoreObject, call the backend to store the data
    immediately, instead of keeping it in RAM or in the temporary object table.
    Data is then referenced only by its checksum.
    Without such change, the storage could fail to store the transaction due to
    lack of RAM, or it could make tpc_finish step very slow.
  - Backends have to store data in a separate space, and remove entries as soon
    as they get unreferenced. So they must have an index of checksums in object
    metadata space. A new '_uncommitted_data' backend attribute keeps references
    of uncommitted data.
  - New methods: _pruneData, _storeData, storeData, unlockData
  - MySQL: change vertical partitioning of 'obj' by having data in a separate
    'data' table instead of using a shortened 'obj_short' table.
  - BTree: data is moved from '_obj' to a new '_data' btree.

- Undo is optimized so that backpointers are not required anymore to fetch data:
  - The checksum of an object is None only when creation is undone.
  - Removed DatabaseManager methods: _getObjectData, _getDataTIDFromData
  - DatabaseManager: move some code from _getDataTID to findUndoTID so that
    _getDataTID only has what's specific to backend.

- Removed because already covered by ZODB tests:
  - neo.tests.storage.testStorageDBTests.StorageDBTests.test__getDataTID
  - neo.tests.storage.testStorageDBTests.StorageDBTests.test__getDataTIDFromData

This changes how NEO stores undo information
and how it is transmitted on the network.

Also clean up deleteObject.

With the change in TransactionManager, there's 1 comparison less.

During 'NEOCluster.stop', there's often 1 'Serialized.background()' call too
many, and when it removed the last lock of the queue, another thread would hang
if was sleeping between 'q.put(lock)' and 'q.get().release()'.

An easy way to reproduce the bug was to add:
    if cls.pending is None:
        time.sleep(.01)
just before 'q.get().release()'