kafka: check reachability + collect metrics

[xvello]

• Make sure capture does not start if the Kafka brokers are unreachable. We'll want to tie that to a readiness probe later instead.
• Collect kafka producer metrics every 10 seconds:

capture_kafka_broker_tx_errors_total{broker="1"} 0
capture_kafka_broker_tx_errors_total{broker="-1"} 0
capture_kafka_broker_rx_errors_total{broker="-1"} 0
capture_kafka_broker_rx_errors_total{broker="1"} 0
capture_kafka_producer_queue_bytes 100000
capture_kafka_producer_queue_bytes_limit 1073741824
capture_kafka_producer_queue_depth_limit 100000
capture_kafka_produce_avg_batch_size_bytes{topic="events_plugin_ingestion"} 2036
capture_kafka_callback_queue_depth 0
capture_kafka_broker_requests_pending{broker="-1"} 0
capture_kafka_broker_requests_pending{broker="1"} 0
capture_kafka_producer_queue_depth 1
capture_kafka_produce_avg_batch_size_events{topic="events_plugin_ingestion"} 1
capture_kafka_broker_responses_awaiting{broker="-1"} 0
capture_kafka_broker_responses_awaiting{broker="1"} 0