Parallel streaming: Support concurrent streams from Amp Server, partitioned by block_range

[fordN]

This PR adds parallel streaming support enabling fast historical data backfills with automatic transition to live streaming.

What's New

Parallel Execution for Historical Data

• Partition large block ranges across multiple workers using ThreadPoolExecutor
• 4-8x speedup for historical loads (scales with worker count)
• Block-based partitioning with automatic or manual partition sizing

Hybrid Mode (Parallel Catchup → Live Streaming)

• Auto-detect current max block and load historical data in parallel
• Seamlessly transition to single-stream continuous mode for live blocks
• Configurable reorg buffer (default: 200 blocks) for safe transition overlap

Usage

Basic parallel historical load:

parallel_config = ParallelConfig(
    num_workers=4,
    table_name='eth_firehose.blocks',
    min_block=0,
    max_block=1_000_000,
    block_column='block_num'
)

results = client.sql(query).load(
    connection='postgres',
    destination='blocks',
    stream=True,
    parallel_config=parallel_config
)

Hybrid mode (parallel → continuous):

parallel_config = ParallelConfig(
    num_workers=4,
    table_name='eth_firehose.blocks',
    min_block=0,
    max_block=None,  # Auto-detect and transition to streaming
    reorg_buffer=200  # Configurable overlap for reorg safety
)

Key Features

• Configurable partition sizes and reorg buffer
• Comprehensive logging and statistics tracking
• Rich metadata in LoadResult for monitoring
• Thread-safe statistics aggregation
• Graceful error handling (workers fail independently)
• Full documentation with usage patterns and troubleshooting
• Create table earlier in flow, so all parallel workers can be sure it's already there before starting

Testing

• 28 unit tests (partitioning logic, query transformation, stats tracking)
• 4 integration tests (parallel load, hybrid mode, block detection)

Documentation

• User guide: docs/parallel_streaming_usage.md - comprehensive usage patterns, configuration options, performance characteristics, and troubleshooting
• Examples: Quick start, hybrid mode, checkpointing, custom partitioning