fix(mail): drop stale Amavis pid files on container restart

[Frtrillo]

Summary

• On docker restart, openship-mail keeps leftover Amavis pid/lock/socket files in /var/run/amavis.
• Net::Server then abort-loops (Pid_file already exists for running process) when that PID now belongs to another daemon (usually Dovecot).
• Originating mail defers on 127.0.0.1:10026 even though Health still shows Amavis as running.

Changes

• apps/email/docker/entrypoint.sh: Unlink stale Amavis runtime files and recreate /var/run/amavis before supervisord starts. This entrypoint is the first process in a fresh pid namespace, so those files cannot refer to a living Amavis.
• apps/email/docker/supervisord.conf: Note why the entrypoint does this, next to the Amavis program.

This matches the existing ClamAV /run handling (issue #565): Debian's tmpfiles.d rule does not run under supervisord, and docker restart is not the same as docker recreate.

Test plan

• Scoped to the mail engine image entrypoint; no API/dashboard behavior change
• Recreate vs restart: docker restart openship-mail should leave Amavis listening on 127.0.0.1:10024 and 127.0.0.1:10026
• supervisorctl status amavis should be RUNNING, not crash-looping on the leftover pid
• Deferred originating mail should flush (postqueue -f) once Amavis is up
• Health → Outbound delivery should not report connect to 127.0.0.1:10026: Connection refused after restart

Verified on a live ghcr.io/oblien/openship-mail:0.7.2 box: leftover pid 67 was Dovecot after restart; removing the pid file and restarting Amavis unstuck 6 deferred messages through SES.

Made with Cursor