fix(node): route domain errors from native EventEmitter

[andrewtdiz]

Fixes #4395.

Summary:

• Route new EventEmitter() / new events.EventEmitter() through the native events handle path used by the no-auto events archive.
• Add external EventEmitter domain hooks so domain.add/remove, ee.domain, and unhandled error emits work across stdlib and native events handles.
• Reserve low native handle IDs from GC heap pointer scanning so tagged native handles are not traced as heap objects.
• Avoid cross-allocator temporary buffers in dynamic handle dispatch and Error stack construction paths reached by the fixture.

Validation:

• PERRY_NO_AUTO_OPTIMIZE=1 target/release/perry compile test-parity/node-suite/domain/members-emitter-errors.ts -o /tmp/perry-domain-members-debug --debug-symbols --no-cache && /tmp/perry-domain-members-debug
• PERRY_NO_AUTO_OPTIMIZE=1 target/release/perry compile test-parity/node-suite/events/imports/default-import.ts -o /tmp/perry-events-default-import --debug-symbols --no-cache && /tmp/perry-events-default-import
• PERRY_NO_AUTO_OPTIMIZE=1 target/release/perry compile test-parity/node-suite/events/imports/named-imports.ts -o /tmp/perry-events-named-imports --debug-symbols --no-cache && /tmp/perry-events-named-imports
• PERRY_NO_AUTO_OPTIMIZE=1 target/release/perry compile test-parity/node-suite/events/imports/namespace-import.ts -o /tmp/perry-events-namespace-import --debug-symbols --no-cache && /tmp/perry-events-namespace-import
• env CARGO_BUILD_JOBS=1 RUSTC_WRAPPER= cargo check -p perry-ext-events -p perry-runtime -p perry-stdlib
• env CARGO_BUILD_JOBS=1 RUSTC_WRAPPER= cargo test -p perry-runtime native_handle -- --nocapture
• env CARGO_BUILD_JOBS=1 RUSTC_WRAPPER= cargo test -p perry-runtime test_write_barrier_non_pointer -- --nocapture
• rustfmt --edition 2021 --check <touched files>
• git diff --check

[proggeramlug]

Superseded by #4398 ("fix(domain): route emitter errors through domains"), which already merged and closed the same issue #4395. main's domain-error routing (incl. the domain_handle field on event emitters) comes from #4398's accepted implementation, so this PR's alternative approach is now redundant and heavily conflicting. Closing.