fast-vad drops Rust, Python VAD

This is the kind of audio tooling developers actually need: small, fast, language-agnostic, and built for real-time pipelines instead of benchmark theater.

• –The project ships usable interfaces for both Rust and Python, which makes it easier to plug into inference servers, speech apps, and research workflows
• –Streaming support is the standout feature because many open-source VADs are easier to batch than to deploy in low-latency systems
• –The design is deliberately simple: 32 ms frames, FFT-derived band features, and a hardcoded logistic-regression classifier with smoothing rules
• –A crate release is already live on Lib.rs, which makes this feel more like a real package launch than a one-off GitHub dump
• –The main open question is benchmarking quality versus established VAD options, but the packaging and speed-first engineering are genuinely compelling