fix: migration deadlock under helm --atomic --wait

[vishnu-narayanan]

Summary

The post-install/post-upgrade migration hook deadlocks under helm upgrade --install --atomic --wait. Web pods can't pass their /health probe before the schema is created (Rails AccountDashboard class-load queries installation_configs), but Helm waits for web Ready before firing the post-install hook. Affects DigitalOcean Marketplace installs (chatwoot/chatwoot#14360) and any user passing --atomic --wait.

Changes

• migrations-job.yaml -> plain batch/v1 Job (no helm.sh/hook annotation), version-suffixed name via new chatwoot.jobSuffix helper, ttlSecondsAfterFinished: 86400.
• web-deployment.yaml + worker-deployment.yaml -> wait-for-migrations initContainer that polls postgres with pg_isready, then waits for the installation_configs table to exist.
• values.yaml -> drop hooks.migrate.hookAnnotation (no longer used).

Pattern mirrors GitLab's official chart: regular Job + wait-for-deps initContainer on app pods.

Validated

Fresh DOKS cluster, all four scenarios pass with --atomic --wait:

• Fresh install with the DO Marketplace command.
• Upgrade from broken 2.0.22 to fixed chart.
• Same-version no-op upgrade.
• Chart-version bump (replays migration with new Job name).