Solution

[dmitryjum]

1. Ruby solution:
   I've created 3 types of AwardsMananger class for benchmarking my ideas mostly. You can see that one (my favorite one) instantiates just one object of AwardsManager, accepts an array of awards as an argument and mutates each of them in the loop of its instance method. The second one instantiates a SingleAwardManager object per every award in an external loop and then mutates it (I expect it to be the least performant solution). And the third one doesn't instantiate NoInstanceAwardManager class at all, but mutates it using its class methods.
   I personally expected the last one to win the competition. And it does most of the times:

All awards update:   0.319682   0.013307   0.332989 (  0.333321)
single instance award update:   0.568429   0.001684   0.570113 (  0.570984)
no manager instantiation award update:   0.318844   0.000713   0.319557 (  0.319983)

But if you run the benchmark a number of times you can find it loose to my favorite solution that accepts an array of awards:

All awards update:   0.322307   0.012791   0.335098 (  0.335346)
single instance award update:   0.547072   0.001719   0.548791 (  0.549390)
no manager instantiation award update:   0.334916   0.001093   0.336009 (  0.337913)

2. Python solution implementation resembles my preferred Ruby AwardsManager implementation.