In particular, a cluster may now starts automatically so 'simple' can't know
whether the cluster is in an expected state or not.

Because masters don't have persistent storage, the task to remember whether
the cluster is in backup mode or not is delegated to storages, via the presence
of a 'backup_tid' config value.

This fixes a bug that set 'backup_tid' after a simple replication.
If the cluster was restarted, it would have tried to switch to BACKINGUP state.

It would be unreliable to resume printing logs using a condition on date,
so the format of log files is changed to add primary keys that number records.

This should fix the following failure:

Traceback (most recent call last):
  File "bin/neostorage", line 23, in <module>
    neo.scripts.neostorage.main()
  File "neoppod/neo/scripts/neostorage.py", line 80, in main
    app.run()
  File "neoppod/neo/storage/app.py", line 167, in run
    logging.exception('Pre-mortem data:')
  File "/usr/lib/python2.6/logging/__init__.py", line 1088, in exception
    self.error(*((msg,) + args), **{'exc_info': 1})
  File "/usr/lib/python2.6/logging/__init__.py", line 1082, in error
    self._log(ERROR, msg, args, **kwargs)
  File "/usr/lib/python2.6/logging/__init__.py", line 1173, in _log
    self.handle(record)
  File "/usr/lib/python2.6/logging/__init__.py", line 1183, in handle
    self.callHandlers(record)
  File "neoppod/neo/lib/logger.py", line 210, in callHandlers
    self._queue(record)
  File "neoppod/neo/lib/logger.py", line 200, in _queue
    self._release()
  File "neoppod/neo/lib/logger.py", line 79, in _release
    self._async.pop()(self)
  File "neoppod/neo/lib/logger.py", line 96, in flush
    self.db.execute("BEGIN")
sqlite3.OperationalError: cannot start a transaction within a transaction

This is a follow up of commit 1f6bbcc1
("Rewrite (inside protocol, to make it standalone) a simpler Enum type"),
which renumbered values of CellStates enums.

In STOPPING state, the primary:
1. rejects all clients and any new node
2. finalize remaining transactions
   (i.e. those for which a tcp_finish was received)
3. finally shut down all nodes of the cluster

During election, nodes are not identified anymore.
testElectionWithManyMasters only worked thanks to the election timeout.

Also, change the way TTIDs are generated in preparation for that bug's fix:
we will need TTID to be monotonous across master restarts and TID generator
provides this feature.

This should never trigger, as gmtime bases itself on a unix epoch
timestamp which is non-monotonous on leap seconds. So gmtime cannot
return second count above 59.

Prepares for type change so it is easily reversible.

Fixes two bugs:
- IPv6 addresses are now put in canonical form.
- non-IPv6 port-less addresses are resolved.
Also, simplify url-style IPv6 parsing.
Also, rename "ip" into more generic naming "host".

The only case where UUID persistence matters is on storage nodes, and it
is already restored from database if available.

Missed in 4df3c4e1.

Since lib.protocol has been rewritten to be semantic-oriented and
duplicates a lot less code, this test is just bloated (testing the same
code over and over) and a maintenance burden.

It is intended to raise even when running in optimised mode. The intent of
that change was to allow interpreter to detect the last elif is not
expected to be needed (although it is not known to make such deductions
currently). If only python had a deeper "enum" notion...
This (very) partially reverts f32e9120.