feat(mqtt): metrics and observability (Dropwizard / Prometheus)

[mkjsix]

Sub-issue of #601.

Summary

Integrate the mqtt module with RESTHeart's existing metrics module to expose MQTT ingestion health via Dropwizard Metrics and the Prometheus /metrics endpoint.

Motivation

Without metrics, operators have no visibility into whether the MQTT bridge is healthy, whether messages are being dropped, or whether the MongoDB write buffer is near capacity. This is essential for production deployments.

Proposed metrics

Metric	Type	Description
mqtt.messages.received	Meter	Messages received from broker per topic
mqtt.messages.dropped.rate-limit	Meter	Messages dropped by global router rate limiter
mqtt.messages.dropped.queue-full	Meter	Messages dropped due to full per-connection SSE queue
mqtt.pipeline.dropped	Meter	Messages discarded by pipeline stages (per stage type)
mqtt.buffer.size	Gauge	Current MongoDB write buffer fill level (0–capacity)
mqtt.buffer.dropped	Meter	Messages dropped by write buffer overflow
mqtt.writer.batch-latency	Timer	Time from message receipt to MongoDB insert
mqtt.writer.dead-letter.count	Counter	Cumulative failed batches written to dead-letter
mqtt.broker.connected	Gauge	1 = connected, 0 = disconnected
mqtt.broker.reconnects	Counter	Cumulative broker reconnect attempts

All metrics are tagged with the topic filter where applicable.

Scope

• MqttMetrics singleton initialized by MqttClientProvider at startup.
• Injected (not via @Inject — internal singleton) into MqttMessageRouter, MqttSseService, and MqttMongoWriter.
• Metrics registered with RESTHeart's MetricRegistry using the mqtt. namespace.
• Health check: mqtt.broker.connected == 0 marks the /health endpoint as degraded.

Dependencies

• Blocked by feat: MQTT client module — bridge external broker events to HTTP/SSE/MongoDB #601.
• Requires the metrics module to be enabled.