Add math/thinking reward

[DNXie]

• Add Reward abstract class
• Add MathReward class
• Add unit tests for MathReward
• Add ThinkingReward class
• Add unit tests for ThinkingReward

Test:

python tests/unit_tests/data/test_math_reward.py

..........................
----------------------------------------------------------------------
Ran 26 tests in 0.002s

OK

python tests/unit_tests/rl/test_thinking_reward.py
................
----------------------------------------------------------------------
Ran 16 tests in 0.000s

OK

[DNXie]

@pbontrager If multiple numbers appear, take the last one. Does this design sound reasonable to you?

[pbontrager]

I think for now we should just have a rewards.py file instead of a whole folder so it's easier to scan through. Otherwise looks good!

[DNXie]

I think for now we should just have a rewards.py file instead of a whole folder so it's easier to scan through. Otherwise looks good!

Solved in #68