[codex] Add scientific validation and live concordance panel

[senseibelbi]

What changed

• added an offline scientific-validation harness that runs orchestrator scenarios and emits JSON/Markdown scorecards with assessment rollups and interop attachment coverage
• added a live CTX concordance reference panel that exercises the real evidence synthesizer against curated ToxVal cases with expected robust/limited outcomes
• extended evidence concordance matching so it can compare against real ToxVal-style fields such as toxicologicalEffect and toxvalNumeric
• documented the new validation layers in the workflow/testing docs and exported the new APIs from the package surface

Why this changed

The orchestrator validation story was still mostly structural. We needed a reproducible way to detect scientific agreement drift in the actual evidence synthesis path, especially when upstream CTX payloads evolve.

Impact

• maintainers can now run scripts/scientific_validation_report.py for offline structural validation
• maintainers can now run scripts/live_concordance_panel.py for live CTX-backed concordance drift detection
• evidence synthesis is materially more defensible against real ToxVal payload shapes

Validation

• pytest -q tests/workflows/test_live_concordance_panel.py tests/workflows/test_scientific_validation_report.py tests/workflows/test_offline_workflows.py
• python3 scripts/live_concordance_panel.py --json
• python3 scripts/scientific_validation_report.py --scenario acute_toxicity --json
• pytest -q