fix: Failed jobs cleanup attempts are recorded as successful and skipped for a full interval

[sam-saffron-jarvis]

What changed

• updated maybeRunCleanup() to record lastCleanupCompleted only after pruneOldData() succeeds
• preserved the existing cleanup interval semantics by storing the cleanup start time on success
• added a regression test that forces cleanup to fail and verifies the next attempt is not suppressed

Why this is high-value

A transient cleanup failure was being treated as a successful cleanup run. That caused the scheduler to skip retention cleanup for the full configured interval, which defeats the existing error-backoff retry path and can let job_runs_v2 / job_run_events_v2 grow unchecked for hours or days. This fix ensures failed cleanup attempts keep retrying on the scheduler error delay until pruning succeeds.

Validation

• gofmt -w cmd/serve_jobs_v2.go cmd/serve_jobs_v2_test.go
• go build ./...
• go test ./...
• git diff --stat
• git diff --check