feat(clone): add structuredClone support for Error objects and relate…

[akshatnathani]

This Pull Request fixes/closes #5279.

It changes the following:

• Added Error object support in the JsValueStore serialization path so structuredClone(new Error("boom")) and postMessage(new Error("boom")) no longer fail during cloning.
• Implemented Error object reconstruction in the deserialization path using stored error metadata (kind, name, message, and related fields), instead of returning Not yet implemented.
• Added regression tests for both clone paths:
  • clone::tests::clones_error_objects
  • message::tests::basic_error_message
• Removed the stale #[expect(unused)] on the ValueStoreInner::Error variant, since it is now actively used.
• Verified consistency with project standards and checks:
  • cargo fmt --all -- --check
  • cargo clippy -p boa_runtime --tests -- -D warnings
  • cargo test -p boa_runtime clones_error_objects -- --nocapture
  • cargo test -p boa_runtime basic_error_message -- --nocapture

[Monti-27]

cause is getting flattened into a string here before it goes into the store. that works for simple text causes, but it breaks real cases like new Error("boom", { cause: { code: 7 } }), where the cloned error should still keep the original object as its cause.

[akshatnathani]

Great catch. I fixed the cause handling so that it is preserved as structured data rather than stringified. Added regression coverage for both structuredClone and postMessage paths, and also looking into the AggregateError issue you pointed out, and will follow up with the patch/tests shortly.

[Monti-27]

this rebuilds AggregateError with an empty list, so the cloned value loses its errors property on the way back out. i checked this with a quick repro and structuredClone(new AggregateError([new Error("inner")], "agg")) comes back with errors.length === 0.