Add integrated scylla-monitoring stack support

[dkropachev]

Summary

Add first-class monitoring integration to CCM: a Prometheus + Grafana + Alertmanager
stack that runs alongside any Scylla cluster as Docker containers with --net=host.

Two modes of operation:

• Automatic (--monitoring flag or CCM_MONITORING=1 env var) — monitoring
  starts with the cluster and Prometheus scrape targets are kept in sync on every
  topology change (add, remove, start, stop, decommission).
• Manual (ccm monitoring start/stop/sync) — on-demand control, no automatic
  target updates.

Key features:

• Per-cluster port offsets (based on cluster ID) for running multiple monitored clusters
  simultaneously
• Auto-clones scylla-monitoring repo for real Grafana dashboards; falls back to a
  built-in overview dashboard when unavailable
• Atomic target file writes so Prometheus never reads partial state
• Monitoring failures in automatic mode are logged as warnings and never block cluster
  operations
• Configuration persisted in cluster.conf and restored on load

New CLI surface:

ccm create ... --monitoring [--monitoring-dir=PATH]
ccm monitoring start|stop|enable|disable|sync|status

Changed files

• ccmlib/scylla_monitoring.py — new MonitoringStack class
• ccmlib/cmds/cluster_cmds.py — ClusterMonitoringCmd + --monitoring flag on create
• ccmlib/cluster.py — monitoring fields, _notify_topology_change() hook
• ccmlib/scylla_cluster.py — auto-start/stop monitoring, CCM_MONITORING env var
• ccmlib/scylla_node.py, ccmlib/node.py — topology change notifications
• ccmlib/cluster_factory.py — restore monitoring settings on load
• docs/monitoring.md — full reference documentation
• README.md — monitoring section, environment variables table
• tests/test_scylla_monitoring.py — unit tests for MonitoringStack
• tests/test_monitoring_integration.py — integration tests for CLI and hooks

[dkropachev]

@fruch , now it is ready and tested, the only reason why ccmlib/scylla_monitoring.py is so big is that there is no way to reuse all the scripts from https://github.com/scylladb/scylla-monitoring, because it is shared environment, there could be many stacks running, so i had to switch to host networking, so provisioning now is manual and scripts have no cli to help with it.