Fix #1741: record calls on stubbed getters/setters

[youdie006]

Purpose (TL;DR)

stub.get(fn) / stub.set(fn) installed the raw replacement function directly as the property accessor, so invoking a stubbed getter/setter ran untracked — callCount never incremented (#1741). This wraps the replacement in a spy and exposes it as stub.getter / stub.setter, so accessor invocations are recorded (mirroring the existing spy.get / spy.set surface). This is the scoped "tiny PR" @fatso83 green-lit in the issue thread.

How to verify

const obj = { get prop() { return 1; } };
const s = sinon.stub(obj, "prop").get(() => 42);
obj.prop; obj.prop;
s.getter.callCount; // 2  (previously: not recorded)

npx mocha test/src/stub-test.js --grep "records calls to the stubbed" → 2 passing.

Checklist

• Lint passes (eslint --max-warnings 0; prettier clean)
• Scoped + non-breaking (+19 lines in default-behaviors.js, no API removed); full suite 1554 passing / 0 failing locally.

Fixes #1741

[mroderick]

This is a really clean fix. Thank you for picking it up, and for keeping it scoped — the diff is easy to follow and exactly addresses the issue.

A couple of nice things I noticed 👏

• Wrapping in spy() is the right approach — routes through the existing call-recording machinery without reinventing anything.
• this context is properly preserved through the spy proxy.
• No circular dependency introduced (spy.js doesn't import default-behaviors.js).
• The naming collision with stub.get/stub.set makes .getter/.setter the right call.

One small suggestion: resetBehavior in stub.js doesn't clean up .getter/.setter, so after a resetBehavior() the stub still holds stale spy references. It's a narrow window (calling .get() again overwrites it), but adding delete this.getter; delete this.setter; there would keep it clean.

Nice work on this!