test(int-factor): complete Phase 3 conformance

[kim-em]

Closes #9627.

Summary

• record independent Phase-2 review attestations for HexIntFactor and HexIntFactorMathlib
• author the missing bridge SPEC, classify the existing Mathlib companion as correspondence-only, and document its computational owners
• complete route, checker, arithmetic, order, cyclotomic, primality-adapter, and divisor-ordering conformance through Phase 3
• make the correspondence-only performance-owner report check phase-aware, with Phase-3/Phase-4 boundary regressions

Phase provenance

The bridge SPEC is newly authored here to make the already-landed correspondence surface and runtime boundary explicit; the counters were lagging existing implementation. A fresh isolated Claude Opus reviewer independently checked that SPEC against the source, transported Mathlib APIs, downstream needs, Phase contracts, and this PR diff. It found no blocker. Its four actionable observations were all addressed: document factorization_absent, clarify the attestation provenance, externally cross-check the ascending divisor enumeration, and add actual prime-table boundary cases.

Verification

• python3 -m unittest scripts.test_check_dag
• python3 scripts/check_dag.py
• structural, phase, release-manifest, trust-surface, and conformance-target checks
• lake build HexIntFactor HexIntFactorMathlib HexConformance
• focused rebuilds after review changes
• fresh hexintfactor_emit_fixtures output matches the committed JSONL byte-for-byte
• python3 -m py_compile scripts/oracle/intfactor_pari.py

The required PARI/python-flint oracle remains wired in the existing single CI job. Local fixture regeneration succeeded; CI supplies the unavailable local cypari2 dependency and performs the required oracle run.