[codex] prove integer refinement overflow parametrically

[briansrls]

Summary

Adds an R3 gate #21 integration receipt proving integer refinement overflow handling is parametric across fixed-width integer refinements, rather than a one-off UInt8 check.

The new test table exercises source-representable out-of-range literals for Int8, Int16, Int32, UInt8, UInt16, UInt32, UInt64, and UInt128, asserting the same structural MagnitudeOutOfRange diagnostic with exact target bounds and no fabricated fixes. It also checks alias routing for the widths whose overflow edges are fully representable by the current i64 source literal carrier.

Validation

• cargo fmt --all --check
• cargo test -p v3-compiler --test integration int_refinement_overflow_is_proven_parametric_for_representable_widths -- --nocapture
• cargo test -p v3-compiler --test integration int_literal_cardinality_test -- --nocapture

[briansrls]

Review metadata

• Provider / model: codex / unknown
• Commit: f69f2ab8 · Trigger: schedule
• Thinking: 248s wall

---

Non-blocking — Strengths

• src/v3/compiler/tests/integration/int_literal_cardinality_test.rs The added implementation-level overflow table checks the shared MagnitudeOutOfRange path across signed, unsigned, alias, and current source-carrier-limited integer refinements without introducing substrate state.

✅ No blocking concerns in the changed Rust-test surface.