Fix crash on cyclic re-export of an imported binding

[littledivy]

Re-exporting an imported binding (import { x } from "m"; export { x }) from a module in an import cycle could dereference a NULL JSVarRef and crash (SIGSEGV). While a peer in the cycle is still unlinked its import slot is NULL, and both js_build_module_ns() and js_inner_module_linking() used that slot directly.

Minimal reproducer (segfaults on master):

// p.mjs
import { q } from "./r.mjs";
export { q as p };
// r.mjs
import { p } from "./p.mjs";
export { p as q };
// entry: import { p } from "./p.mjs";

js_get_local_export_var_ref() resolves such a binding by following the import alias to its source module, reusing js_resolve_export1's JSResolveState visited-set to detect re-export cycles. When the binding is genuinely unresolvable (a pure re-export cycle with no concrete binding) it throws a SyntaxError instead of crashing.

case	before	after
valid cyclic re-export (#567)	stopgap SyntaxError	evaluates correctly
pure re-export cycle, no binding	crash	SyntaxError
mutual re-export of imports	crash	resolved or SyntaxError

This also lets the test_cyclic_import.js case that carried a FIXME(bnoordhuis) work as intended. Fixes #567.

Disclaimer: This is an AI-assisted patch from the v8x project.

[saghul]

While you're here, can you please rename the test to bugXXX.js ?

[littledivy]

Renamed to bug567.js