fix: #837 — emit wrapper for default/null/undefined-named exports

[proggeramlug]

Summary

• The wrapper-emission site at crates/perry-codegen/src/codegen.rs:2552-2586 keyed __perry_wrap_<...> on each function's local HIR name, while the consumer-side reference at crates/perry-codegen/src/expr.rs:~3819 builds the same symbol from the exported name. For renamed exports — export { local as exported } or export default function NAME — the exported wrapper symbol was never defined and the link failed.
• A new emission pass right after the regular wrapper loop walks hir.exports, picks out Named { local, exported } entries with exported != local, and emits __perry_wrap_perry_fn_<src>__<sanitize(exported)> with external linkage.
  • Function rename → forwards to perry_fn_<src>__<local> using the closure-call ABI.
  • Variable / class / type rename, or export default function NAME whose body the HIR lowerer never recorded (crates/perry-hir/src/lower.rs:5296 TODO) → no-op returning undefined, mirroring the existing __perry_wrap_perry_unknown_func fallback.

Reproductions (pre-fix)

#837 — uuid's dist/esm-browser/v35.js uses export default function v35(...). Consumers in v3.js/v5.js pass v35 as a closure value, which transitively pulls the wrapper symbol:

Undefined symbols for architecture arm64:
  ___perry_wrap_perry_fn_node_modules_uuid_dist_esm_browser_v35_js__default, referenced from:
      ___perry_wrap_perry_fn_node_modules_uuid_dist_esm_browser_v3_js__v3
      ___perry_wrap_perry_fn_node_modules_uuid_dist_esm_browser_v5_js__v5

#836 (partial) — zod's src/v4/core/regexes.ts uses const _null = /.../; export { _null as null }; (same for undefined). A consumer that reads regexes.null / regexes.undefined as a value falls through to the function-shape branch and references:

___perry_wrap_perry_fn_node_modules_zod_src_v4_core_regexes_ts__null
___perry_wrap_perry_fn_node_modules_zod_src_v4_core_regexes_ts__undefined

Validation

• Post-fix nm on node_modules_uuid_dist_esm_browser_v35_js.o now shows T ___perry_wrap_perry_fn_<v35.js>__default. uuid repro reaches the linker successfully (sha1.js's Call callee shape not supported (PropertyGet) with 20 args is a separate codegen gap and surfaces independently; the wrapper-emission failure no longer blocks).
• Post-fix nm on node_modules_zod_src_v4_core_regexes_ts.o now shows T ___perry_wrap_..._null and T ___perry_wrap_..._undefined. The repro program (import * as regexes from "zod/v4/core/regexes"; console.log(typeof regexes.null);) links and runs, printing function (the closure handle, per js_closure_alloc_singleton semantics).
• cargo build --release clean.
• cargo test --release --workspace --exclude perry-ui-ios --exclude perry-ui-tvos --exclude perry-ui-watchos --exclude perry-ui-visionos --exclude perry-ui-android --exclude perry-ui-windows --exclude perry-ui-gtk4 — all green (no regressions in the 231-test perry-codegen / perry suites or anywhere else in the workspace).
• cargo fmt --all clean.

Scope notes

• The uuid v35.js runtime ("v3() returns undefined" rather than a uuid string) is a separate HIR-level bug: crates/perry-hir/src/lower.rs:5296 has a literal // TODO: properly lower function expression in the ExportDefaultDecl::Fn arm and never registers the function body in module.functions. This PR's scope is codegen-only; the runtime fix needs an HIR change and a different PR.
• #836's cross-module class re-exports are out of scope per the task brief — that's a separate PR.

Refs #837 (link error only — runtime gaps tracked in #871).
Refs #836 (partial — null/undefined wrappers fixed; cross-module class re-exports remain a separate PR).

Test plan

• Reproduce both link failures pre-fix (see above).
• Confirm __perry_wrap_..._default is emitted with external linkage in v35.js's .o.
• Confirm __perry_wrap_..._null / __perry_wrap_..._undefined are emitted in regexes.ts's .o.
• Confirm uuid repro reaches linker (separate sha1.js codegen issue surfaces post-fix; no regression).
• Confirm zod regex repro compiles and runs.
• Workspace cargo test clean.