fix(docker): add metadb postgres dependency to mc-infra-connector (fixes clean install Wave 1 failure)

[MZC-CSC]

Problem

On a clean install, installAll.sh fails at Wave 1:

mc-infra-connector exited with code 2 (restarting)
mc-infra-manager | ... GET mc-infra-connector:1024/spider/readyz ... lookup mc-infra-connector on 127.0.0.11:53: server misbehaving
mc-infra-manager | ... setup: CB-Spider not ready, retrying (17/60)
dependency failed to start: container mc-infra-manager is unhealthy
❌ Service startup failed (exit code: 1) - Wave 1 (mc-infra-connector mc-infra-manager ...)

Root cause

• Since cb-spider 0.13.9 (Add missing configuration for infra connector #191), its metadata DB runs as a separate container, mc-infra-connector-postgres.
• mc-infra-connector has no depends_on for this container.
• The startup waves in installAll.sh list only entry-point services. They expect compose to start everything else through the depends_on chain, so Wave 1 never starts the postgres container.
• cb-spider cannot resolve mc-infra-connector-postgres:5432 and exits with code 2 in a restart loop. mc-infra-manager then becomes unhealthy while waiting on /spider/readyz, and Wave 1 fails.
• Existing installs don't hit this because postgres is already running. It only reproduces on a clean install.

A related gap: MC_INFRA_CONNECTOR_POSTGRES_HOST_PORT was missing from .env.setup. The port mapping resolved to :5432, so postgres was published on a random host port.

Changes

• conf/docker/docker-compose.yaml: add depends_on: mc-infra-connector-postgres (condition: service_healthy) to mc-infra-connector. The postgres service already has a pg_isready healthcheck.
• conf/docker/.env.setup: add MC_INFRA_CONNECTOR_POSTGRES_HOST_PORT=6433, next to the infra-manager postgres on 6432. For existing installs, sync_missing_env_vars in installAll.sh adds it to .env automatically.
• docker-compose.mini.yaml does not need this change. It still uses cb-spider 0.12.18 with a SQLite meta_db.