Accept WebAssembly.Module values in the dynamic worker loader

[guybedford]

This implements support for passing compiled WebAssembly.Module objects into the dynamic worker loader modules bag, allowing compiled Wasm modules to be shared with loaded workers without recompilation.

The motivating flow is source phase imports into dynamic workers:

import source mod from './lib.wasm';

const worker = env.LOADER.load({
  compatibilityDate: "2026-08-13",
  mainModule: "index.js",
  modules: {
    "index.js": `
      import source mod from './lib.wasm';
      export default { /* ... WebAssembly.instantiate(mod) ... */ };
    `,
    "lib.wasm": mod,
  },
});

What was implemented:

• The worker loader modules bag accepts a WebAssembly.Module as a module value, or { wasm: module } alongside the existing { wasm: bytes } form.
• WorkerSource::WasmModule carries an optional v8::CompiledWasmModule; both the legacy and new module registries use FromCompiledModule when present (the new registry seeds its existing compile cache), so the loaded worker shares compiled code with zero recompilation. The loader and loaded worker are always in the same process, which is exactly the boundary v8::CompiledWasmModule supports; the module wire bytes remain available as the module body.
• jsg::V8Ref<v8::WasmModuleObject> is now unwrappable, with rtti and TypeScript type updates (Record<string, string | WebAssembly.Module | WorkerLoaderModule>).

Test coverage: end-to-end loader tests (source phase import → loader → import source + default import in the child, against both module registries, covering both accepted forms).

In future, dynamic import() could work directly against these module objects as well, per the ESM Phase Imports proposal.

[codspeed-hq]

Merging this PR will not alter performance

✅ 72 untouched benchmarks
⏩ 129 skipped benchmarks¹

---

_{Comparing gbedford/wasm-module-serialization (caf8765) with main (b6424d1)}

Footnotes

1. 129 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[jasnell]

I know these are fine but they do require a double take to verify they are safe... at first glance it looks like bytes ends up dangling. Maybe worth expanding the comment more.