chore(example): do not report malformed notification responses as success

[OskarEichler]

Description

One Line Summary

Require an object response and let JSON parsing failures reach the existing failure handler. Do not retry ambiguous malformed responses.

Compatibility and observable changes

Example only. Invalid success bodies report failure without another send, avoiding a duplicate notification when delivery status is unknown. Existing explicitly recognized subscription-indexing retries remain unchanged.

Details

Motivation

The source audit reproduced the failure paths described below. This PR contains only the associated fix; unrelated audit changes are in separate PRs.

Scope

• examples/demo/src/services/OneSignalApiService.ts

Testing

Mocked-fetch diagnostics cover null, non-object, malformed JSON and valid response controls; no notification is sent to the OneSignal service.

Each code/tooling fix was also applied independently to upstream commit a70312207cf094ac361eaa9196c317acb175c2cd and passed its targeted external actual-source diagnostics. Documentation snippets were checked separately. Native diagnostic harnesses use bridge/SDK doubles and are not an end-to-end push test.

On the combined audit branch:

• Existing SDK suite: 5 files, 262 tests pass; unchanged 95% coverage thresholds pass.
• vp check: formatting, lint and type checks pass; native Spotless check passes.
• Both example apps: Android Debug and unsigned iOS Simulator builds pass.
• Both example apps: iOS and Android production Metro bundles pass.

No checked-in test files were added or modified; regression evidence comes from external diagnostic harnesses and the existing suite. No physical-device, live notification delivery, Appium/BrowserStack, or release-workflow execution is claimed. The no-location example's stale native lock was updated locally to resolve the current SDK for verification; generated locks are not part of this PR.

Checklist

• Required description sections completed.
• Scope and observable/API behavior explained.
• Diff reviewed and targeted regression checks run.
• Automated checks and device-testing limitations documented.