Add zero-dependency parallel redaction for large inputs

[DBinK]

Summary

• Add an opt-in parallel feature with Redactor::find_parallel and Redactor::clean_parallel.
• Use scoped standard library threads, leave one available CPU for other work, and keep inputs smaller than 256 KiB on the serial path.
• Preserve zero dependencies, no_std + alloc support, detector priority, masking behavior, and panic propagation.
• Use BTreeMap for overlap resolution to avoid quadratic insertion costs on match-dense inputs.
• Add serial and parallel equivalence tests, a dense-match complexity guard, documentation, and a dependency-free benchmark comparison.

Performance

Command: cargo run --release --all-features --example bench

Synthetic 1,420,090-byte input on the development machine:

• Serial: 16.4 MiB/s
• Parallel: 52.0 MiB/s
• Speedup: 3.17x

Testing

• cargo fmt --all -- --check
• cargo clippy --all-targets -- -D warnings
• cargo clippy --all-targets --all-features -- -D warnings
• cargo test --all-features
• cargo build --no-default-features
• cargo doc --no-deps
• cargo test --release --test complexity -- --ignored --nocapture
• cargo publish --dry-run --allow-dirty
• CodeRabbit review: 0 findings after the final review

Notes

The local Rust 1.70.0 toolchain download failed during the TLS handshake with static.rust-lang.org, so the MSRV-specific Clippy and test commands were not run locally. The pull request CI matrix covers Rust 1.70.0.

[ptukovar]

Thank you