feat(bigquery-jdbc): implement non-blocking telemetry batcher and dispatcher

[Neenu1995]

Finalizes the TelemetryBatcher infrastructure for high-performance, non-blocking telemetry collection and dispatching in the BigQuery JDBC driver.

Key Functionality Added:

• Non-Blocking Telemetry Ingestion: Buffers connection attempts, statement executions, error metrics, and feature usage events in capacity-bounded LinkedBlockingQueue storage (max 10,000 items per queue).
• Fine-Grained Concurrency Control: Uses an explicit ReentrantLock (flushLock) for thread-safe flush operations, task rescheduling, and shutdown without locking producer threads.
• Efficient Payload Sizing: Implements single-pass bulk payload trimming to strictly enforce the 512 KB payload threshold before transport transmission.
• Optimized Batch Draining: Leverages BlockingQueue.drainTo() to drain telemetry batches in a single operation, minimizing lock acquisition overhead.
• Conditional Resource Allocation: Lazily allocates the ScheduledExecutorService and background dispatcher thread only when telemetry is explicitly enabled.

[gemini-code-assist]

Code Review

This pull request introduces TelemetryBatcher, a class designed to buffer and periodically flush telemetry events, and increases the default batch size threshold to 5000. The review feedback highlights critical performance bottlenecks in the implementation. Specifically, it recommends replacing ConcurrentLinkedQueue with LinkedBlockingQueue to avoid O(N) size() operations, refactoring helper methods to use the generic Queue interface, and optimizing the payload trimming loop in flush() to remove items in bulk rather than one-by-one to prevent expensive, repetitive protobuf serialization size checks. Additionally, using explicit locks instead of synchronized blocks is recommended for performance-sensitive code.

[Neenu1995]

/gemini review

[gemini-code-assist]

Code Review

This pull request introduces TelemetryBatcher to buffer and periodically flush telemetry events, along with increasing the default batch size threshold in TelemetryConfiguration from 100 to 5000. The review feedback focuses on improving robustness and performance, suggesting: avoiding potential IllegalStateException by using queue.offer instead of queue.addAll when requeuing; optimizing queue draining by declaring queues as LinkedBlockingQueue and using drainTo to reduce lock contention; and lazily initializing the ScheduledExecutorService only when telemetry is enabled to prevent resource leaks.