Test scheduled-post job outcomes with fakes

[brudex]

Proposed labels: testing, area:backend, help wanted

Context

Scheduled-post cron modules make status, schedule-status, and batch-counter
transitions that are important to publishing reliability. The root test suite
currently concentrates on API contracts and release smoke checks rather than
these job outcome combinations.

Suggested starting points

• job-runner/schedule-due-posts-cron.js
• job-runner/publish-scheduled-cron.js
• job-runner/publish-recurring-cron.js
• job-runner/lib/publish-post-accounts.js
• test/

Scope

Add the smallest practical dependency seams and fake models/publishers needed
to cover no-due-post, all-success, partial-success, and total-failure paths.
Assert post status, schedule status, and relevant batch counters for each path.

Non-goals

• Do not change scheduling or retry policy unless a failing test demonstrates
  an existing defect and the maintainer agrees on the behavior.
• Do not invoke real providers, cron daemons, or database infrastructure.

Acceptance criteria

• Tests cover all four outcome classes for the affected scheduled/recurring
  flow.
• Assertions include final post and schedule status plus batch accounting where
  the module maintains it.
• The test suite is deterministic and makes no network calls.
• Production behavior is unchanged unless a separately documented bug fix is
  included with a regression test.

Verification

1. Run npm test twice to confirm deterministic results.
2. Confirm all provider/model boundaries in the new tests are faked.
3. Run git diff --check.