Fix duplicate completion registration in generated progress promises

[leileizhang (lei9444)]

Problem

Generated JavaScript async-with-progress values already start and project a native Promise. Calling their public .toPromise() registered completion again, so a stock InMemoryRandomAccessStream.writeAsync(buffer).toPromise() rejected with 0x80000018 (E_ILLEGAL_DELEGATE_ASSIGNMENT) instead of returning the awaited result.

Fix

Return the existing projected Promise from .toPromise(), including the existing rejection for pre-aborted signals. The shared emitter covers operations, actions, and overloads. Clarify this identity guarantee without changing signatures, raw DynWinRtValue.toPromise(), Python, native ABI, or runtime dispatch.

Hosted Rust 1.99 exposed pre-existing formatting drift in two test-hook stub macro calls. Accept optional trailing commas and format those calls so the normal CI lane can run; emitted native signatures and behavior are unchanged.

Verification

• Fresh exact-main (de91e9b3) output fails the new 4 MiB stock stream regression with 0x80000018 on ARM64 and x64. Fresh repaired output passes on both using locally built, matching production runtimes (no preview binaries/test hooks).
• The ABI-free execution double measures one completion registration/result projection, action undefined results, overload dispatch, repeated/detached .toPromise(), await/.then()/Promise.all, projected progress, cancellation, rejection identity, listener cleanup, and pre-aborted signals with zero native calls. Strict TypeScript consumer regression passes.
• Stock stream write/read and owned-localhost HTTP progress/cancel/abort/error regressions run in isolated, 30-second-bounded children through the normal E2E specs/schema/runner and CI. HTTP overload/content/buffer consumers also check projected object identity and bigint progress, adding 196 covered lines to the unchanged fixture set; coverage thresholds/exclusions are untouched.
• Local checks: new regression tests (2), related snapshots/declaration tests (10), codegen unit suite (650 passed, 1 ignored), standard generated JavaScript E2E (39/39 plus 4 implementation scenarios), existing raw async/progress tests (5), and Rust formatting with both 1.97 and CI-matched 1.99. Native test-hook compilation passes on ARM64/x64; E2E preparation contract tests pass.

Independent PR based directly on main; no unmerged #198/#205 or audit-branch changes.

[github-actions]

Mixed-language test coverage

Workflow status: ✅ Passed

Layer	Lines	Functions	Branches/regions
Rust, including native .pyd/.node	86.97%	82.49%	86.7% regions
Python aggregate	72.34%	n/a	39.42% branches
Python runtime	98.17%	n/a	94.57% branches
Generated Python WinRT projections	71.07%	n/a	33.5% branches
Generated Python WinRT implementations	71.97%	n/a	46.3% branches
JavaScript aggregate	18.18%	24.41%	56.01% branches
JavaScript runtime	44.27%	45.76%	78.99% branches
Generated WinRT projections	16.59%	18.52%	52.17% branches
Generated WinRT implementations	45.99%	59.19%	60.97% branches
Generated Classic COM projections	11.88%	23.97%	53.4% branches

View workflow run and download full HTML/LCOV/XML reports