Optimize online difficulty filtering logic

[thanay-sisir]

🚀 Summary

Refactors the difficulty filtering mechanism in the rollout buffer to improve performance and maintainability. Replaces chained conditional logic with a dictionary-based lookup, reducing complexity from O(n) to O(1).

🛠 Key Changes

• Refactor: Replaced multiple if/elif comparisons with a direct difficulty_mapping dictionary lookup.
• Logic:
  • 0.0 reward → hard
  • 1.0 reward → easy
  • Others → normal
• Extensibility: Centralized mapping logic makes future difficulty granularities easier to implement.

⚡ Performance Impact

• Time Complexity: Constant time O(1) lookup.
• Efficiency: Reduces CPU cycles and branch prediction misses during high-frequency rollout processing.

[samsja]

I don't think that the extra logic is usefull / needed. Thanks for your contribution tho