feat: fix Prometheus metrics + add health probes and log aggregation (observability)

[NotYuSheng]

Description

The observability stack is incomplete and partly broken. Identified in the production readiness assessment (#366, docs/production-readiness.md, findings P1-1, P1-2, P1-6).

1. Prometheus metrics do not work

application.yml:81-93 exposes a prometheus actuator endpoint and enables the Prometheus registry, but backend/pom.xml declares only spring-boot-starter-actuator — the micrometer-registry-prometheus dependency is missing, so /actuator/prometheus returns 404 and nothing is scrapeable. Additionally, nginx does not proxy /actuator (nginx/nginx.conf.template), so actuator is only reachable on the backend container's :8080.

2. No health probes

The backend and nginx containers define no Docker healthcheck (only PostgreSQL and MinIO do). There are no liveness/readiness probes. management.endpoint.health.show-details: when-authorized plus no security means health details are never shown.

3. No log aggregation

All services log to stdout (Docker-captured only). The prod profile would write a rotating file to /var/log/tracepcap/ but no volume is mounted for it and the profile is inactive (see #366 P0-3).

Acceptance Criteria

• Add micrometer-registry-prometheus; /actuator/prometheus returns metrics.
• Expose/scrape metrics in the target deployment (Kubernetes ServiceMonitor).
• Add liveness/readiness probes hitting /actuator/health for backend (and a healthcheck for nginx).
• Ship container stdout to a log aggregation stack (e.g. Loki/ELK); drop the unmounted file-logging path or mount a volume for it.

Affected Files

• backend/pom.xml
• backend/src/main/resources/application.yml
• nginx/nginx.conf.template
• deployment manifests (see research: production readiness assessment #366 Helm work)