feat: Add Prometheus metrics endpoint

[initcron]

Summary

Add a /metrics endpoint that exposes gateway metrics in Prometheus format for production observability.

Parent Epic

Part of #1 - Production Kubernetes & Container Support

Proposed Metrics

Gateway Metrics

Metric	Type	Description
mcp_gateway_requests_total	Counter	Total MCP requests by server_id, method
mcp_gateway_request_duration_seconds	Histogram	Request latency histogram
mcp_gateway_active_servers	Gauge	Currently running server instances
mcp_gateway_server_starts_total	Counter	Server startup count by server_id
mcp_gateway_server_stops_total	Counter	Server shutdown count by server_id, reason
mcp_gateway_errors_total	Counter	Error count by server_id, error_type

Server-Level Metrics

Metric	Type	Description
mcp_server_tools_called_total	Counter	Tool invocations by server_id, tool_name
mcp_server_uptime_seconds	Gauge	Time since server started
mcp_server_memory_bytes	Gauge	Memory usage (if available)

Implementation

Dependencies

[dependencies]
prometheus = "0.13"

Endpoint

GET /metrics
Content-Type: text/plain; version=0.0.4

Example Output

# HELP mcp_gateway_requests_total Total MCP requests
# TYPE mcp_gateway_requests_total counter
mcp_gateway_requests_total{server_id="postgres",method="tools/call"} 142

# HELP mcp_gateway_active_servers Currently running servers
# TYPE mcp_gateway_active_servers gauge
mcp_gateway_active_servers 3

Acceptance Criteria

• /metrics endpoint returns Prometheus-format metrics
• Request count and latency tracked per server
• Active server count gauge
• Error rate tracking
• Compatible with Prometheus scraping
• Optional: ServiceMonitor CRD for K8s

References

• Prometheus Rust client
• Prometheus exposition formats