Add Tiny Aya text generation option

Settings: recommended quality presets + device readout; persist all choices

Model bar: show params once (avoid '0.6B · 0.6B' for byte-less engines)

Fix model re-downloads (persistent storage) + WebLLM shader-f16 crash

Multi-engine benchmark: wllama vs Transformers.js vs WebLLM (engine selector)

Add Qwen3-0.6B; surface WebGPU backend; strip <think> from answers

In-browser models: catalog + picker + cache delete, live tok/s stats, raw 'thinking' stream