perf: optimize range API with streaming & prefetch

[notTanveer]

Summary

• Streaming range endpoint: Replace full-buffer approach with chunked byte streaming and Cache-Control headers.
• Block prefetch: Prefetch the next block in both Bitcoin Core and Esplora providers to hide RPC/HTTP latency during sequential scans
• Binary precomputation: Precompute and store silent block data in binary format at index time, eliminating per-request serialization
• WebSocket streaming sync: Add filterSpent support and batched binary framing to the WS gateway for efficient wallet sync

Benchmark Results (tested with shroud)

Architecture	50-Block Sync Time	Client Throughput	Payload per 1,000 Blocks
V1: REST JSON (filterSpent=true)	~6.00s	~2,500 tx/sec	~20 MB
V2: REST Binary (Unfiltered)	~1.50s	~16,700 tx/sec	~58 MB
V3: REST Binary (filterSpent=true)	~2.00s	~9,500 tx/sec *	~12 MB
V4: WebSocket Binary Streaming (Current)	< 0.50s	~20,000 tx/sec	~12 MB

Key Wins

1. Network Pipelining: WebSocket streaming eliminates the HTTP overhead and constant polling latency.
2. Maximum Wasm Efficiency: The Rust WebAssembly engine is now parsing heavily filtered UTXOs at a blistering 20,000 transactions per second. Batch scanning consistently finishes in ~570ms for massive 1.5MB chunks.
3. Bandwidth Reduction: Implementing filterSpent=true strictly on the binary endpoint reduced sync payload size by ~80%, avoiding the Base64/JSON serialization tax entirely.