Task: Standardize backend failure injection controls for tests

[RtlZeroMemory]

Parent EPIC: #324

Problem:

Backend failure behavior is currently covered in a fragmented way and needs reusable controls.

Scope:

• Define reusable backend-side failure controls for requestFrame, pollEvents, and lifecycle edge cases.
• Make the controls deterministic and runner-friendly.
• Avoid embedding one-off failure logic in scenario tests.

Done when:

• Scenario and integration tests can inject backend failures through a supported API.
• Backend failure tests no longer need bespoke mock logic for common failure classes.

Notes:

• This task should produce work that is reviewable against expected behavior, not current implementation shape.
• Keep outputs straight to the point and tied to repo code or repo process.

[RtlZeroMemory]

Implemented in #367.

Added reusable shared-backend failure queues in packages/core/src/app/__tests__/stubBackend.ts for start, pollEvents, and getCaps.

This now pins:

• startup failures can be injected without one-off test backends
• poll failures fault the app and still dispose the backend cleanly
• capability lookup failures can drive startup fallback tests through the shared stub path

Verified with:

• ./node_modules/.bin/tsc -b packages/core/tsconfig.json packages/node/tsconfig.json --pretty false
• node --test packages/core/dist/app/__tests__/backendFailureControls.test.js packages/core/dist/app/__tests__/eventPump.test.js packages/core/dist/app/__tests__/terminalProfile.test.js

Still unresolved:

• stop/dispose failure injection remains limited to current shared runtime contracts; no broader lifecycle fault API was added without stronger evidence