Detect shared worker RNG with identical-parameter tasks

To test whether a simulator's own random functions (initial-condition rand(), white-noise terms) repeat across multiprocessing.Pool workers, give every task the same parameters, pass chunksize=1, and hash both the initial state and one recorded trajectory per task, along with os.getpid(). If workers share the RNG state, the count of distinct hashes equals the largest number of tasks any single worker ran (e.g. 3 of 12 on 4 workers), and sorting by pid shows the same hash sequence in every worker. Run three controls: spawn, python -m module on the 3.14 forkserver (the forkserver preloads __main__ only when it is a file), and a reseed at task start. Also note that a new Pool forked from an unchanged parent replays the same streams. A loop that opens one Pool per condition reuses the same noise for every condition.