feat: add comprehensive health check endpoints for server and plugins

[reldothescribe]

Summary

This PR adds comprehensive health check endpoints for the MCP server and plugins.

Changes

New Health Endpoints

1. /health - Overall server health

   • Returns server status, version, and timestamp
   • HTTP 200 with JSON response: {"status": "healthy", "version": "...", "timestamp": "..."}

2. /health/ready - Readiness probe (server initialized)

   • Returns "ready" when the server is running and initialized
   • Returns HTTP 503 if the server is not ready

3. /health/live - Liveness probe (server running)

   • Returns "alive" when the server is running
   • Returns HTTP 503 if the server is not alive

4. /health/plugins - Per-plugin health status

   • Returns overall status and per-plugin health check results
   • Includes plugin status, message, and timestamp for each plugin

Plugin HealthCheck Interface

• Added HealthCheck() method to the Plugin interface in pkg/plugin/plugin.go
• Added HealthStatus type with constants: healthy, unhealthy, unknown
• Added HealthCheckResult struct with status, message, and checked_at fields

Plugin Implementations

Implemented HealthCheck() for all existing plugins:

• clickhouse: Reports number of clusters configured, checks proxy datasources
• prometheus: Reports number of instances configured
• loki: Reports number of instances configured
• dora: Reports number of networks with Dora explorers available

Registry Updates

• Added HealthChecks() method to plugin.Registry to aggregate health from all plugins

Backward Compatibility

• Legacy /ready endpoint is preserved for backward compatibility

Tests

• Added comprehensive tests for all health check endpoints
• Added tests for plugin registry health checks

Testing

Run the tests:

make test

Related

This implementation follows Kubernetes health probe conventions and provides a foundation for monitoring the MCP server in production environments.