Improve Russian Roulette

[SeverinDiederichs]

Two issues were left open by #545:

• Currently, it does exactly what Athena does. However, there is a potential issue in case the neutron Russian Roulette and gamma RR have different weights. In that case, the RR could be applied twice (and then, incorrectly) to particles.
  The behaviour in Athena will be fixed with this MR.
  Solved in AdePT by use simple unity check in ATLAS Russian Roulette #598
• A CI test should be put in place to test the RR. This can be done by adding a testEm3 like geometry, which has the same naming scheme as ATLAS uses to check for RR, with a "LAr" prefix.
  Solved in AdePT by add full CI test and drift test for Russian Roulette #599