XY-899: [ELF benchmark vNext P3] Add qmd and OpenViking strength-profile benchmarks

[yvette-carlisle]

Summary

• Add qmd/OpenViking strength-profile benchmark report artifacts with scenario-level outcome boundaries.
• Preserve qmd debug/replay ergonomics as not-tested where no equivalent scored ELF surface exists.
• Keep OpenViking trajectory strengths as research-gate/not_encoded behind typed wrong_result same-corpus evidence.
• Align external adapter project-count semantics across generator, spec, tests, and measurement coverage audit.

Validation

• cargo test -p elf-eval --test real_world_job_benchmark qmd_openviking_strength_profile_report_preserves_claim_boundaries --all-features
• cargo test -p elf-eval --test real_world_job_benchmark real_world_report_includes_external_adapter_coverage_manifest --all-features
• cargo test -p elf-eval --test real_world_job_benchmark current_benchmark_reports_preserve_live_sweep_boundaries --all-features
• cargo make real-world-memory-live-adapters
• ELF_BASELINE_PROJECTS=qmd,OpenViking cargo make baseline-live-docker
• cargo make fmt
• cargo make lint-fix
• cargo make checks

Notes

• Scoped baseline run exited 0; qmd passed and OpenViking remained typed wrong_result evidence, not a loosened pass.
• Independent Decodex review checkpoint is clean for head 154bdbf.