v6.0.5

NDL optimization: Benchmark Total Time (each test = hundreds of thousands of operations):

Before (Linear Search):
10m: ████████████████████████████████████████ 18.98s (263k ops × 72µs each)
20m: ████████████████████ 9.31s (540k ops × 17µs each)
30m: ████████ 3.73s (1.3M ops × 3µs each)
40m: █████ 2.41s (2.1M ops × 1µs each)

After (Binary Search):
10m: █████ 2.46s (-87.2%) ⚡⚡⚡ (263k ops × 9µs each)
20m: █████ 2.71s (-71.1%) ⚡⚡ (540k ops × 5µs each)
30m: █████ 2.72s (-26.9%) ⚡ (1.3M ops × 2µs each)
40m: █████ 2.47s (-5.0%) ✓ (2.1M ops × 1µs each)

Benchmark Environment

• Tool: Criterion.rs v0.5.1
• Samples per benchmark: 100
• Total operations: 263k - 2.1M (varies by depth)
• Warmup: 3 seconds per benchmark
• Measurement: 5+ seconds per benchmark
• Compiler: rustc 1.83.0 (release mode)
• Platform: macOS (Apple Silicon)
• Confidence: 95% (p < 0.05)

What's Changed

• cns and otu moved to deco model trait by @KG32 in #53
• ndl optimization - binary search by @KG32 in #54

Full Changelog: v6.0.3...v6.0.5