second instead of some fixed number. Keeps benchmark faster by putting a cap on total execution time. Before a run using importlib took longer by some factor, but now it takes roughly the same amount of time as using the built-in __import__.