Parallel time-range sync for 4-5x speedup

[kittleik]

What

New parallel_sync.py script that splits the sync time range into N segments and syncs them in parallel using ThreadPoolExecutor.

Key features

• Parallel workers: Splits (last_timestamp → now) into N segments, each synced independently
• Sticky cursor optimization: Skips tiny follow-up batches when boundary count < 100 (was ~50% of all batches)
• Graceful shutdown: SIGINT/SIGTERM handling, preserves temp files
• Per-worker logging: Each worker logs to parallel_logs/worker_N.log
• Safe merge: Temp files merged into main CSV in timestamp order after all workers complete

Test results

• 2 workers, ~30 min range: 36,380 records in 28.6s (1,273 rec/s)
• Original sequential: ~1,000 rec/s
• With 5 workers on the full 50-day gap: expect 4-5x throughput

Usage

pkill -f update_goldsky.py  # stop current sync first
python3 parallel_sync.py --workers 5