fix(electron): normalize hashed standalone externals

[tianrking]

Summary

Fixes the Electron packaging path that left Turbopack-hashed external imports such as ws-a972e7ffa40ff725 in server instrumentation chunks.

assembleStandalone already has patchTurbopackChunks for this exact normalization and the npm prepublish flow enables it. The Electron staging flow did not, so the packaged app copied canonical node_modules/ws but attempted to import the non-existent hashed name at startup.

This enables the existing normalization step for Electron builds and adds regression coverage for both the package-name rewrite and the Electron assembly wiring.

Validation

• node --import tsx/esm --test --test-name-pattern "patchTurbopackChunks restores canonical external package names|electron standalone assembly normalizes Turbopack hashed external imports" tests/unit/build/assemble-standalone.test.ts tests/unit/electron-packaging.test.ts
• npx prettier --check scripts/build/prepare-electron-standalone.mjs tests/unit/build/assemble-standalone.test.ts tests/unit/electron-packaging.test.ts
• npx eslint tests/unit/build/assemble-standalone.test.ts tests/unit/electron-packaging.test.ts --no-error-on-unmatched-pattern --suppressions-location config/quality/eslint-suppressions.json
• git diff --check

The complete assemble-standalone.test.ts suite has an existing Windows-only path failure in its unrelated standalone-server-ws.mjs test: it derives a file: URL with .pathname, producing F:\\F:\\.... The selected new regression and Electron packaging test both pass.

Scope

No provider behavior, route handling, or bundled dependency list changes. The staging step now rewrites only the erroneous hash suffixes before electron-builder copies the standalone bundle.

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

[diegosouzapw]

Thanks for tracking this down — this lines up exactly with #7346 (macOS Electron ws-a972e7ffa40ff725 ERR_MODULE_NOT_FOUND). Good catch that patchTurbopackChunks was already wired for the npm prepublish path but never enabled for the Electron staging step; reusing the existing, already-tested normalization rather than writing a new one is the right call here.

I verified locally in an isolated worktree: both new regression tests pass, the full assemble-standalone.test.ts and electron-packaging.test.ts suites pass, eslint is clean, and confirmed this doesn't touch the non-Electron standalone path (build-next-isolated.mjs keeps the flag off by default) or overlap with the v3.8.48 win-spawn/better-sqlite3 rebuild hotfix.

One tiny nit (non-blocking, won't hold up the merge): the inventory-table comment at the top of assembleStandalone.mjs (around line 33) still lists the Electron column for the Turbopack hashed-chunk patch as "-" — worth a follow-up doc tweak to flip it to "Y" now that this lands, just for future readers of that table.

Planning to merge this as-is. Will also link it to close #7346.

[diegosouzapw]

Merged into release/v3.8.49 — thank you for the contribution, @tianrking! 🎉 Validated in today's full-suite merge-train (33 PRs, 19k+ tests green) before landing.