weavatrix-scan 0.4.3

• Adds path-free ScanSummary aggregation for full and compact reports.
• Speeds ignore-aware scanning by reusing prepared directory rules, batching ignore-source discovery, avoiding successful-path skip allocations, and prefiltering complex globs.
• Shares full and compact parallel batch processing without changing their public report contracts.
• Rebalances shallow traversal across eight workers; the pinned Rust raw row improves from 421.1 ms to 285.9 ms.
• Adds fixed Rust and Gitea corpus benchmarks, dirwalk and std::fs::read_dir controls, and exact-output parity gates.
• Keeps TypeScript tooling out of the crate; serde_json and dirwalk remain development-only dependencies.

On the pinned Windows Rust corpus, compact/full parallel scans measured 4.77x/3.81x faster than ignore. On Gitea they measured 5.63x/5.38x faster. Raw traversal remains an explicit gap: dirwalk measured 123.3 ms versus 285.9 ms for collected ParallelWalker.