Port decisiongrounding onto the shared harness (scored-result equivalence)

[tcballard]

Second-phase initiative, sequenced after the restructure move landed (decisiongrounding/ arrives unchanged per the frozen repo-topology item — this issue must not expand that item's scope).

Recorded finding from scoping: decisiongrounding is a scenario/provider benchmark with a structural adherence scorer (scoring/scorer.py), pre-registered specs, and append-only results; it has no {metrics, baseline} scorecard or committed query set today, so the planned "capture the current scorecard metrics block as the equivalence reference" step has no surface to capture yet. The port therefore needs, in order:

1. Define its equivalence surface (a metrics projection of its deterministic run results, or a frozen report shape).
2. Capture that reference on the unchanged benchmark.
3. Port to consume harness/ (drop the private scorer glue where it duplicates harness responsibilities; keep its corpora, scenarios, and rac/decisions/).
4. Acceptance: post-port scored results equal the pre-port reference on the same inputs; its gate stays green; DG-ADR-0001 still holds.

Roadmap: rac/roadmaps/benchmarks/tool-benchmarks.md in rac-core [roadmap:tool-benchmarks].