feat(runtime,codegen): Object.getOwnPropertyDescriptors (#1791/#1758)

[proggeramlug]

Summary

Implements Object.getOwnPropertyDescriptors(obj) (the plural form) — the
first blocker stopping import * as S from "effect/Schema" from initializing
(epic #1785 / #1772, capstone #1791 / #1758).

The singular Object.getOwnPropertyDescriptor(obj, key) was already wired end
to end, but the plural had no dedicated HIR variant. It fell through both
the literal-callee recogniser (lower/expr_call/native_module.rs) and the
esbuild-alias recogniser (destructuring/var_decl.rs), so codegen lowered the
callee to a constant 0.0 (null) and the call threw
TypeError: value is not a function.

effect's SchemaAST.annotations clones an AST node during Schema.ts module
init via:

Object.create(Object.getPrototypeOf(ast), Object.getOwnPropertyDescriptors(ast))

so the import threw before any user code ran.

Localization

PERRY_DEBUG_SYMBOLS=1 + lldb (break js_throw_type_error_not_a_function) +
--trace llvm pinned the throw to AST.annotations (SchemaAST closure
__182), where the callee lowered to %r5 = bitcast double 0.0 to i64 →
js_closure_call1(null, ast). Minimal repro: Object.getOwnPropertyDescriptors({a:1}).

Changes (mirror the singular-descriptor path)

• runtime js_object_get_own_property_descriptors (object_ops.rs):
  enumerate own keys via getOwnPropertyNames, build a fresh object mapping
  each key to its descriptor (js_object_get_own_property_descriptor).
  Non-rooted builder allocations follow the existing js_object_entries /
  getOwnPropertyNames convention.
• new HIR Expr::ObjectGetOwnPropertyDescriptors(Box<Expr>); lowered in both
  the direct-call and Object-alias paths; whitelisted in the alias gate.
• codegen arm (misc_methods.rs) + extern decl (runtime_decls/strings.rs) +
  the codegen dispatch list, the ref/mut walkers, a stable_hash tag, and the
  JS emitter (perry-codegen-js).

Validation

• Byte-for-byte with node --experimental-strip-types across
  plain / empty / alias / accessor-descriptor / clone-via-Object.create shapes
  (new gap test test_gap_object_get_own_property_descriptors.ts).
• effect/Effect deep import stays green (result: 42).
• Class-expr epic gap suite (identity / static-this / extends-static /
  inherited-static-method / inherited-rest-static / new-instanceof /
  static-iife / static-symbol): all pass.
• Object-method tests (define-property, getPrototypeOf, indirect-define-property
  Object.defineProperty with TypeError: value is not a function #886, accessor Object.defineProperty accessor descriptors: getter this binding broken (returns NaN) #450, etc.): all pass.
• 0 regressions. test_issue_685 remains the documented pre-existing
  class-object static-field value-read known-failure (params type: string vs
  object), unrelated to this change.

Remaining (reported, not chained here)

Full effect/Schema init hits a distinct second blocker once this lands:
TypeError: Cannot read properties of undefined (reading '_tag') deeper in
Schema.ts init.

Backtrace:

js_throw_type_error_property_access
  perry_closure_..._SchemaAST_ts__239 (+39352, a ~5.8k-line recursive AST dispatcher)
  perry_closure_..._Schema_ts__211
  perry_closure_..._Schema_ts__227
  node_modules_..._Schema_ts__init_body

The simple class-object static-field value-read this looked like (the
deferred #1787 / #685 sub-item) reproduces correctly in isolation now
(make(x).ast._tag matches node), so the _tag-undefined is a more specific
data-flow issue in Schema's struct construction — to be localized in a
follow-up.

Refs #1791, #1758, #1785, #1772.