Add simdjson chunked parse + domain rewards optimization

[jorgecuesta]

Summary

• simdjson integration for large block_results parsing: Settlement blocks (1.5GB+, 6-9M events) that previously crashed the indexer now parse in ~40s using a custom C++ addon with chunked JSON.parse
• Domain-based rewards computation: Added domain rewards tracking and optimized DB batch operations
• Refactored modToAcctTransfers handling: Improved efficiency of module-to-account transfer processing

simdjson Details

• Added vendor/simdjson submodule (fork of simdjson_nodejs with Buffer support + findChunkBoundaries)
• Two-path strategy: <500MB uses V8 JSON.parse, >=500MB uses simdjson chunked parse
• No lazyParse/valueForKeyPath (segfaults on Node 18 with >1GB buffers)
• Improved HttpClient logging: UTC timestamps, request height, full timing breakdown (headers/body/concat/parse)
• Production verified on mainnet: block_results@679773 total=199s (headers=77s body=80s concat=2s parse=40s) size=1542.8MB

Test plan

• Docker build passes (Node 18 Alpine)
• simdjson loads and works in container
• Normal blocks parse correctly (<500MB fast path)
• Settlement blocks parse correctly (>500MB chunked path)
• Data equivalence verified: all 6.7M events match original parser output
• Deployed to mainnet and beta — running stable
• Benchmarks in benchmarks/json-parse/