Add Distributed Locking to Prevent Race Conditions in Pipeline Declaration

[butschster]

When RoadRunner starts its worker pool for the jobs plugin, each worker process attempts to declare consumer pipelines during initialization. This creates a race condition where multiple workers try to create the same pipeline simultaneously, potentially leading to conflicts, duplicate creation attempts, and errors.

This PR introduces a distributed locking mechanism to the RPCPipelineRegistry class to ensure that only one worker process can declare a specific pipeline at a time during the worker pool initialization phase.

This issue is particularly problematic for the Kafka driver, where each pipeline declaration creates a Kafka consumer that joins the consumer group. When multiple workers declare the same pipeline simultaneously, phantom consumers are created in the Kafka consumer groups. These phantom consumers remain idle and don't consume messages from their assigned partitions, while only one consumer actually processes messages. This leads to:

• Unprocessed messages sitting in partitions assigned to phantom consumers
• Reduced throughput as available partitions are not being consumed