Out-of-gate emit census: binder types, A4 order on rust fixtures, refinement kind capture

[gunbai-bot]

Summary

• Census (snappy-stag-806, pin 9bb7440, 223/920) reported silent reds on out-of-gate emit/infer claims. Re-derived per DESIGN §6b; these three share signature-encoding drift, not a gate-admission change.
• Ledger (row, cause, disposition, this PR)
  • produced_item_parameter_order_test an_item_emits_its_parameters_in_declared_order / a_domain_canonicalized_item_emits_the_same_parameter_order — (b) claim drifted from XL-2 binder nodes. Renderer params are binder_edges; the claim still treated e.target as a type atom. Fixed the claim to read declared_field_from_edge (still discriminates [z:Int, a:Bool] vs stored [a,z]).
  • rust_literal_bind_emit_test bool/char (int already used dag_arrow_with_body_node) — (b) fixture drifted from A4: rust_arrow_with_body_node had no ^arrow_signature_order_edge, so translate_arrow_signature_split refuses. Now uses declared_signature like dag_arrow_with_body_node.
  • module_member_emission_test a_where_refinement_emits_its_transparent_alias_on_the_production_route_holds — (a) kind capture used a production-shell walk while lowering classifies refinements with body_lower_find_where_host_conj. Capture now uses that finder so type Pos = Int where positive is DeclaredRefinement on the production route.
• Not in this PR (classified, left): infer_expected_type_record_instantiation_* is enrolled expected-red (checking mode). infer_generic_formal_is_counted_frontier / rcf_a_cross_module_* still need a main re-run. function_value_body_route is calm-boar-904. Do not add these modules to the required gate.

Test plan

• CI floor/generated on this head
• After merge-queue pressure drops: claim_batch --entry the three modules above
• Confirm bool/char literal emit still discriminates wrong values; where-refinement still must not emit struct

Made with Cursor

[gunbai-bot]

Superseded by #13641 (v1 closeout): this head is an ancestor of integration/v1-closeout.