cjs: find-my-way@9 init throws 'Cannot convert undefined or null to object' — blocks #6559 e2e

[proggeramlug]

Found while landing #6559 (PR #6584): find-my-way@9 compiles natively (compilePackages) but throws TypeError: Cannot convert undefined or null to object at load — before any route is registered or any runtime-generated code runs.

Uncharacterized beyond the failure point: something during module init passes undefined/null into an Object.* operation (likely Object.keys/assign/setPrototypeOf over a value perry resolves differently at CJS init — needs a bisect into find-my-way's init graph: handler-storage maps, http-methods table, or its Object.create(null) trees).

Repro: the #[ignore]d e2e in crates/perry/tests/issue_6559_real_libs_e2e.rs (PERRY_REQUIRE_NPM_E2E=1 … -- --ignored).

Blocks: #6559 real-library acceptance (router leg), kimi-code target #6565 (fastify routing = find-my-way). Interpreter side already proven against find-my-way's captured matcher codegen incl. new NullObject() on a host class (PR #6584 fixtures).

[proggeramlug]

Resolved by #6598 (merged to main as 49db1ac).

Root cause: OrdinaryHasInstance (ECMA-262 7.3.21) with a null/undefined left operand threw TypeError: Cannot convert undefined or null to object instead of returning false. find-my-way@9's FindMyWay(opts) is called without new, so its Router body evaluates this instanceof Router with this === undefined; because the RHS is a function value, it routed through js_instanceof_dynamic's prototype-walk tail (ordinary_has_instance_prototype_walk), which called getPrototypeOf(undefined) and aborted module init before any route was registered.

Fix: the prototype walk now short-circuits to false for every non-object LHS (null/undefined/boolean/int32/string/bigint/symbol), per step 3 ("If Type(O) is not Object, return false"), and Symbol() instanceof Object is now false. Verified byte-identical to Node across 26 instanceof cases; real_find_my_way_router in crates/perry/tests/issue_6559_real_libs_e2e.rs is un-ignored and green.

Note: one minor CodeRabbit suggestion — set PERRY_REQUIRE_NPM_E2E=1 for the e2e-scoped CI job so that e2e can't silently skip — was not included in #6598 and remains as optional CI hardening.