Add fermion exchange logic and tests for bhabha scattering

[AntonReinhard]

Currently only tested for n-Photon Compton processes (where it changes nothing) and Bhabha scattering (e- e+ -> e- e+).

Fixes #27

[AntonReinhard]

Okay I found that this does work for n=1 and n=2 fermion lines, but stops working beyond that, as I feared/expected.

[AntonReinhard]

I implemented a new algorithm where the VirtualParticles now also carry information about the remaining open cycles of their subdiagrams. Negation then happens on Pair and Triples tasks, once for every cycle that closes. Because the cayley distances of permutations can be calculated by n-number_of_cycles in a permutation, this comes out to the same when we only care about whether an even or odd amount of transactions are necessary.
I will write some actual documentation on how all of this works at some point, too. And also, hopefully, some more formal proof that it is correct.