[nvdorman]

Summary

• make explicit antares serve detach by default and return after the child owns the configured listener and health is ready
• add antares stop with graceful SIGTERM, bounded wait, and process-group SIGKILL escalation
• add antares status with PID, dashboard URL, version, and daemon log path
• preserve antares serve --foreground and bare headless antares for systemd, containers, and debugging
• store atomic PID identity state under ANTARES_HOME and logs at ~/.antares/logs/daemon.log
• reject stale/reused PID files using kernel start time, UID, executable, command line, and socket ownership
• adopt a legacy foreground antares serve listener so the first antares stop works after upgrade
• serialize concurrent starts and reject occupied-port false readiness

CLI

antares serve               # starts in background and returns
antares status              # reports daemon status
antares stop                # graceful bounded shutdown
antares serve --foreground  # old attached behavior

Verification

• make check
• go test -race ./cmd/antares
• real isolated lifecycle: start, health, status, duplicate start, stop, idempotent stop
• occupied-port startup rejection with no stale PID state
• Linux /proc listener-owner, UID, start-time, and upgraded/deleted executable handling
• Darwin cross-build for foreground portability
• read-only adoption test against a live pre-feature Antares instance