[app] Align cron lifecycle with deployment#72 (Recreate, foreground crond, console lock/watch)

[henzigo]

Context

Upstream PR shopsys/deployment#72 refined the graceful cron lifecycle beyond what phase 1 of
this chart implements. Differences worth adopting:

1. strategy: Recreate — our cron Deployment kept the legacy RollingUpdate
   (maxSurge: 1, maxUnavailable: 0), which briefly runs two cron pods concurrently
   during the date-label-forced replacement. Recreate guarantees the old pod drains fully
   before the new one starts.
2. crond in the foreground (exec crond -f || exec cron -f, log tail backgrounded) —
   proper PID 1 signal handling; our current command backgrounds crond and tails a FIFO as
   the main process.
3. Console commands instead of phing in the preStop drain:
   bin/console deploy:cron:lock / deploy:cron:watch (new application interface) instead of
   phing -S cron-lock / cron-watch; lock output redirected to a file, || true guards.
4. terminationGracePeriodSeconds: 3000 upstream vs our 3600 — align the default (pick one,
   document).

Tasks

• strategy: Recreate for the cron Deployment
• Foreground crond command + backgrounded log tail
• preStop via bin/console deploy:cron:lock/watch (confirm minimum project-base version
  that ships these commands; keep phing fallback configurable if needed)
• Align the grace period default
• Update cron_test.yaml, golden snapshots, deviations doc
• Re-check the cron-suspend hook interplay (scale-to-0 + Recreate)

References: shopsys/deployment#72; related: #3 (supercronic), #19 (native CronJobs — if that
lands, parts of this become obsolete)