feat(scoring): add enterprise issue provenance adapters

[seonghobae]

Summary

Advances #404 with the first reviewable enterprise-domain prerequisite:

• adds factory-sealed, content-addressed EnterpriseSourceRecord, EvidenceSpanRecord, and CounterevidenceRecord adapters under the existing fast_mlsirm.scoring namespace;
• distinguishes direct facts, supported inferences, counterevidence, unresolved ambiguity, and stakeholder value judgments without introducing sentiment or priority arithmetic;
• projects exact spans into the existing shared EvidenceReference contract instead of creating a parallel observation/evidence schema;
• retains no source text and reuses the bounded immutable metadata, descriptive identifier, structured error, and canonical fingerprint boundaries;
• adds deterministic contract/security tests, public documentation, APA 7th primary-source traceability, and an authoritative changelog fragment.

Explicit boundaries

This slice performs no semantic extraction, psychometric calibration, ranking, utility/VOI arithmetic, causal inference, or high-stakes automation. AtomicIssueRecord, StakeholderPerspective, CandidateIntervention, scoring-request compilation, Rust-backed measurement, decision-support arithmetic, and queue routing remain follow-up slices.

Research and standards

• W3C PROV-DM is used as the domain-agnostic provenance reference; the package does not claim to emit a complete PROV serialization.
• NIST AI RMF 1.0 is used for documentation, traceability, human-oversight, and limitation boundaries; no conformity claim is made.
• Current CPython dataclass documentation was checked through Context7 for frozen dataclass and InitVar semantics.
• Consensus search was attempted for peer-reviewed provenance/audit evidence, but the connected monthly search quota was exhausted; this slice therefore makes no literature-dependent empirical claim.

Validation target

• 100% statement and branch coverage for added code;
• complete public docstrings;
• exact-head lint, typing, security, packaging, and changelog render parity;
• independent review before merge.

[coderabbitai]

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro Plus

Run ID: 02b1fec8-b4ba-48e0-bdca-43b51ce9846f

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

_{Comment @coderabbitai help to get the list of available commands.}

[seonghobae]

@jules Please validate the exact current head without expanding scope. Run the repository formatter/linter, the focused enterprise-issue tests with statement and branch coverage, the authoritative changelog renderer (python scripts/render_changelog_fragments.py --update CHANGELOG.md), and the normal CI-equivalent checks available to you. Commit only objective fixes needed for formatting, typing, callback-safe validation, deterministic serialization, coverage, or changelog render parity. Preserve the existing shared scoring contracts and the explicit no-sentiment/no-ranking/no-causality boundaries.

[seonghobae]

@google-labs-jules Please validate the exact current head without expanding scope. Run the repository formatter/linter, focused enterprise-issue tests with statement and branch coverage, python scripts/render_changelog_fragments.py --update CHANGELOG.md, and available CI-equivalent checks. Commit only objective formatting, typing, callback-safety, deterministic-serialization, coverage, source-span-bounds, or changelog-render fixes; preserve the shared scoring contracts and the no-sentiment/no-ranking/no-causality boundaries.

[seonghobae]

Closing as superseded by PR #530, which was opened first from the same base and implements the same fast_mlsirm.scoring.enterprise_issue boundary with the required atomic issue, stakeholder, intervention, and shared scoring-request compilation contracts. Keeping both would create competing schemas and duplicate review/CI effort. Any valid provenance or source-span hardening identified here should be applied to #530 instead.

[seonghobae]

Closing as superseded by merged PR #529 (d78e8854fbdf05c189ba77726e3f36e867f21436). This branch was created from the pre-#529 base and modifies the same fast_mlsirm.scoring.enterprise_issue contracts, but implements only the source/evidence/counterevidence subset already incorporated into the accepted shared schema. Merging or rebasing it would reintroduce a competing contract variant and conflict with the repository's no-duplicate-schema policy. Any genuinely new follow-up should target current main and extend the merged contracts incrementally rather than replace them.