The probabilities comparison in the Bisimulation engine.

[ivan.zapreev]

Somewhere in probabilistic bisimulation we test that probability values are equal. At present this is done up to some error margin, smth like 1e-6. I believe that this parameter has to be available for the user to manage. At least it should be clearly visible that we do probability comparison up to certain error. Also this has to be states in the manual. I suggest:

1. make this parameter a manageable MRMC runtime option
2. Update the MRMC manual correspondingly.

[davidjansen]

The file src/lumping/sort.c currently contains the relevant constant EPSILON, which is #defined to be 3e-12.