Why does execution time of the same algorithm changes so much when it is coded inside a module in numpy

Question

In my learning process, I am writing a value function iteration problem in two different ways.

1. All code inside a python file