Add tests for MixedBehaviorProfile regrets

[rahulsavani]

This sub-issue is the part of issue #444 that relates only to MixedBehaviorProfiles and not MixedStrategyProfiles -- this has been split off as it will be dealt with as part of PR #642, issue #617 (with tests for both the agent and non-agent versions of max_regret for example).

[Shivika-5236]

Hi @rahulsavani @tturocy sir, I’d like to work on this issue. Please let me know if I can proceed with it.

[satyam102006]

Hi @rahulsavani, I would love to pick this issue up!

I have already drafted the tests for both the standard max_regret() and agent_max_regret() methods for MixedBehaviorProfile. I made sure to follow the existing test structure to ensure coverage for both the rational=True and rational=False flags. Everything is passing locally on my end.

For a bit of context on my background, I actively contribute to the OpenCV project (handling C++ core regressions, memory leaks, and bug fixes), so I am very comfortable navigating C++/Python bindings and writing rigorous test coverage.

I will open a PR with these changes shortly

[rahulsavani]

closed via #642