bf_trueF2.py (search over actual power-cell partitions with the true functional F, N <= 6) was started with 5 worker processes and
stopped by hand after about 11 minutes of CPU time per worker, before any job had finished (the nested optimisation is very slow).
It produced no result, and nothing in the paper depends on it: the lower bound is proved for all partitions, and the end-to-end test
main_checks/numerics/true_F_partition_test.py evaluates the true functional on 364 actual partitions.
