Looking at the local results of the new (not yet published) SPOT fast simulation job defined in ATLINFR-4920 I noticed a few performance bottlenecks in TFCSEnergyAndHitGAN. This MR addresses those and is practically identical to !61247 (merged). In a nutshell, we go from 38% CPU utilization w/ 0.091 events/second to 96% CPU utilization w/ 0.453 events/second for 400 events in 8 threads.