XY-932: Add live operator-debug benchmark scoring

[yvette-carlisle]

Summary

• Add Docker-scoped live operator-debug adapter runner for ELF and qmd.
• Add selected-but-not-narrated operator-debug fixture and scored manifest/report rows.
• Update benchmark reports to distinguish the narrow ELF/qmd operator-debug slice from broad viewer-product claims.

Validation

• bash -n scripts/real-world-operator-debug-live-adapters.sh
• jq empty on updated manifest/report JSON and new fixture
• git diff --check
• cargo test -p elf-eval --test real_world_job_benchmark operator_debug_live_adapter_task_is_docker_scoped -- --exact
• cargo test -p elf-eval --test real_world_job_benchmark --all-features

Decodex Recovery

This PR resumes retained partial progress from XY-932 attempt 1 after Decodex stopped during review handoff with app_server_idle_timeout. Runtime artifacts were not committed.