v2: carry the type-declaration modifier slot on NormalizedTree (M0 precursor)

[gunbai-bot]

Precursor to M0 of the nominal-type plan (gunbc#13024, docs/plans/nominal-type-declaration-plan.md). sharp-raven-357 ruled option (2): M0's census rides census-resolve, and that walk could not see whether a type declaration carries nominal_opaque because G0 consumed the modifier slot and minted nothing. Without this change the census's Opaque class would fall silently into UnconstrainedNominal / TrueAlias.

Change

• Vocabulary (v2.compiler.body_lowering_fold): a closed TypeDeclModifier vocabulary (NominalOpaqueModifier | SoleConstructorModifier). It is modeled as the slot plus a closed set, not a per-spelling flag, because plan §9 S1 puts constructor and view visibility in this same slot. The reader body_lower_type_decl_declared_modifiers reads the slot positionally off the parse, the same way body_lower_type_decl_eq_rhs_optional reads the body. A slot it cannot read refuses with body_lowering_reason_type_decl_modifier_unreadable at its location; it is never read as empty.
• Carrier (v2.compiler.normalize): type_declaration_modifier_capture carries the rows on NormalizedTree type_declaration_modifiers, beside type_declaration_kinds, following the DeclaredTypeKind precedent. admit_normalized_tree takes the new argument. Hand-built roots pass no_type_declaration_modifiers(), and roots rebuilt from a tree pass that tree's own rows.
• Consumers: nothing reads the carrier yet, so no existing consumer's behavior changes. The named consumer is M0 (gunbc.instruments.type_declaration_use_census), which lands next (§3c declared frontier).

Rung drop: narrowed, not retired

gunbc.rung_drop.g0_type_decl_modifier_parse_without_sealing_property is narrowed, not retired. This change mints the property, but its restoration trigger is the construction wall, which nothing on the native route enforces yet. A drop is retired by its trigger and by nothing else, so the population row now names the carried-but-unenforced state.

Controls (v2.test.parse.type_decl_modifier_carrier), all PASS locally on the seed

• An unmodified alias carries no row (control).
• nominal_opaque alias carries its modifier, and a sole_constructor record carries its own. Both were red against the first cut of the reader, which mis-read a present keyword atom as an empty slot.
• A generic declaration carries both modifiers in slot order.
• A modifier spelling used as the type's name is not a modifier.
• Pairing claim: the real normalize route carries the row onto NormalizedTree, so deleting the capture turns it red.
• An unknown spelling in the slot still refuses parse_g0_tokens_remain at its location. That is covered by the existing type_decl_modifier_g0_parse_probe (13/13 PASS).

Regression spot-checks of edited call sites (local)

• fold_encoding_test: all PASS.
• image_occurrence_projection_test: all PASS.
• body_let_annotation_test: 34 PASS, 2 FAIL. The same 2 (bla_refinement_*) fail on main at d8eebae, so they predate this change.

Grammar: no production changed; only the comment above dag_grammar_type_decl_modifiers_expr (stern-bear-500 informed).

🤖 Generated with Claude Code