Runs every generation-pipeline process at increasing ticket counts
(100 K → 1 M in 100 K steps). Each process is stopped early when its
throughput drops below 50 % of the 100 K baseline. A PDF report with
per-process scaling curves is generated at the end.