Add BestOfK and Iterative explainer orchestrators

[CharlesR-W]

Summary

• BestOfKOrchestrator: Generates K candidate explanations in parallel, evaluates each with configured scorers, and selects the best explanation based on scorer results.
• HillClimbingOrchestrator: Iteratively refines explanations over multiple rounds using scorer feedback (false positives/negatives) to improve explanation quality.
• Added duration fields to ExplainerResult and ScorerResult for timing instrumentation.
• Added config fields for bestofk_* and iterative_* parameters.

Both orchestrators are standalone classes (no inheritance from Explainer base) that run scorers internally, invoked via if-branches in __main__.py when --explainer bestofk or --explainer iterative is specified.

Test plan

• Verify syntax with python -m py_compile on new files
• Run with --explainer default to confirm no regression
• Run with --explainer bestofk to test best-of-K selection
• Run with --explainer iterative to test hill-climbing refinement

🤖 Generated with Claude Code

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}