Add long run time to player classifier

[drvinceknight]

See chat on #689, idea is to add long_run_time to the classifier dict. We could then change:

long_run_time_strategies = [MetaMajority, MetaMinority, MetaWinner,
                   MetaMajorityMemoryOne, MetaWinnerMemoryOne,
                   MetaMajorityFiniteMemory, MetaWinnerFiniteMemory,
                   MetaMajorityLongMemory, MetaWinnerLongMemory, MetaMixer]

to be

long_run_time_strategies = [s for s in all_strategies if not s.classifier['long_run_time']]

This could be False by default so we'd only need to actually add it to the long_run_time_strategies. All the tests are in place that should pick it up for the lists and we could add a test to the global test player class too...

[marcharper]

We should define long-run strategies somehow. I suggest we pick one strategy as the threshold, and I nominate PSO Gambler.

[drvinceknight]

Haha: I literally just pushed this using the same classification we had already in place: the meta ones. Easy enough to change though: I'm happy with whatever threshold.

[marcharper]

That's where it is already.

[drvinceknight]

Ok cool :)

If we wanted to change it, we just need to change this test and then hunt them down: /~https://github.com/Axelrod-Python/Axelrod/blob/master/axelrod/tests/unit/test_classification.py#L168