📊 Improve Grafana dashboards and observability docs

[deus-developer]

Problem
The provided Grafana dashboards:

• Assume container_name=~".*api" which doesn't match all deployments
• Don't show logs unless Loki is carefully configured
• May duplicate metrics in horizontally scaled setups

Proposal

• Update dashboards to avoid hardcoded filters
• Document how to:
  • Set up traceable logs in Loki
  • Avoid metric duplication from multi-replica services
  • Customize dashboards for non-standard service names

Motivation
Improve out-of-the-box observability. Many users don’t have time to debug Grafana quirks during adoption.