Incorrect calculation of decentralized simulations total running time

[kabirkbr]

Decentralized offernet simulation runs as follows (see code):

1. Runs searchAndConnect routine asynchronously on behalf of each agent in the network;
2. After some time, starts to run cycleSearch routine on behalf of taskAgent, which is known to be part of the cycle;
3. Simulation exists:
   -- when all searchAndConnect routines exit;
   -- AND cycleSearch finds a cycle;
   -- OR timeout is reached (currently one hour);

Yet, sometimes this setup returns more than one cycle (up to 130!). This happens because in a decentralized case a cycle can be found even though not all searchAndConnect routines are finished (due to probabilistic effects related to graph structure). Precisely these effects may make decentralized search advantageous to centralized, which does not consider graph structure.

All this puts negative bias on decentralized search.