Fix a MC serialization bug around WaitAny
The individual wait activities that are embeeded in a WaitAny or
TestAny do not have any call_location associated, as the call_location
is associated to the observer, not the activity. The previous code was
thus not serializing the call location of all those activities while
serializing them, but it's a pity because the Checker code deserialize
these call_location in any case. This was resulting in the beginning
of the next transition to be used as location, naturally breaking the
deserialization afterward.
Instead, the call location of the TestAny or WaitAny is now used for
each and every wait activities embedded in the *Any call. This may
challenge the MC implementation, but it sounds like a sensible idea:
In such case, the location of the embedded wait IS the one of the
englobing TestAny or WaitAny call.
In some sense, that's very close to how we handle the timeout that is
given by the englobing call, not the embedded ones, in such situation.
