Rework Prometheus

[pehala]

• Separate wait for reconciliation
  • Increase timeout to 350 second
• Make Prometheus a Lifecycle object
  • Move client stuff to commit
• Add default Interval for MetricsEndpoint
  • To better aim new wait_for_scrape backoff
• Use ServiceMonitor as base for methods instead of service_url
  • I found it more natural as that is the actual metric definition