Fix diversity check by excluding loss outcome

[AshladBP]

Summary

This PR updates the simulation diversity check to avoid evaluating the loss outcome as part of the "most frequent outcome" diversity threshold.

Background

The project enforces:

• a hit rate constraint (hit rate must be between 5% and 33%)
• a simulation diversity check, warning when the most frequent outcome exceeds 1%

When the loss outcome is included in the diversity evaluation, the warning becomes effectively unavoidable in valid cases, since:

• loss_rate = 1 - hit_rate
• with hit_rate = 5% -33% -> loss_rate = 67% - 95%

So the loss outcome will naturally dominate frequency distribution.

Change

• If the most frequent outcome is loss, the diversity check now evaluates the next most frequent non-loss outcome instead.
• The loss outcome is still shown in results, only the diversity threshold evaluation skips it.

Benefit

This avoids systematic false warnings while preserving the goal of the diversity check: detecting cases where a specific non loss outcome becomes overly dominant.