Fix graph crash with partial conditions

[Synss]

Graph.rep() assumes that every callable has a __name__ attribute.
However, callables like functools.partial and other classes with a
__call__() method have no __name__. This patch modifies
Graph.rep() to first check whether the condition is a string and
return it, then try to return f.__name__ (that is the current
behavior) and if that fails, then it calls str() on the argument.

str() should never fail and it is now the responsibility of the user
to return something meaningful from __str__().

[coveralls]

Coverage decreased (-0.07%) to 99.778% when pulling 1ca302e on Synss:graphing-nameless-callable into 9f7cbe3 on pytransitions:master.

[Synss]

If you like, I can explicitly handle functools.partial and return "check(True)" for

def check(result):
    return result

... conditions=partial(check, True)...

[aleneum]

Partials for condition checks haven't been tested anywhere else yet. If you call fly and make sure that both conditions are checked this would also test if partials are handled correctly. Just in case someone introduces something later on which breaks this compatibility. Like:

# [...]
m = self.machine_cls(states=['A', 'B', 'C'], initial= 'A', show_conditions=True,
                     title='a test')
m.add_state({'name': 'E'})
m.add_transition(trigger='fly', source='A', dest='B',
                 conditions=Check(False))
m.add_transition(trigger='fly', source='A', dest='C',
                 unless=functools.partial(check, False))
m.fly()
self.assertTrue(m.is_C())
# [...]

[Synss]

I could test partial conditions in test_core, make Graph.rep() a staticmethod (it does not use self anyway) and test Graph.rep() directly. Is that OK?

[aleneum]

I could test partial conditions in test_core, make Graph.rep() a staticmethod (it does not use self anyway) and test Graph.rep() directly. Is that OK?

Sounds good. Currently, the 'coverage problem' is, that test_anonymous_callable_conditions does not actually makes use of Check and the partial function check though (see coverage report). So yeah, testing partials for condition checks in core does make sense as well as making rep a module function (does not even have to be part of Graph) but it does not solve the issue of unused constructs in the test itself.
I know its not a big issue but I would prefer if this could be solved in a meaningful/contributing way.

[coveralls]

Coverage decreased (-1.5%) to 98.32% when pulling ef9f4df on Synss:graphing-nameless-callable into 9f7cbe3 on pytransitions:master.

[Synss]

Tests passed on my machine with 2.7.13, I'll try again...

[coveralls]

Coverage decreased (-0.2%) to 99.673% when pulling 235bfa0 on Synss:graphing-nameless-callable into 9f7cbe3 on pytransitions:master.

[aleneum]

Looks good to me, one function name was misspelled, thats why the test hadnt been executed.
Thanks!