FCPE WASM + ONNX Browser Smoke Test

HQ Mode: What Changes?

• Core model is the same. Pitch inference still uses the same FCPE ONNX model and same mel frontend.
• WASM bundle is larger. HQ includes extra decode/ensemble utilities used for advanced post-processing APIs.
• Startup can feel slower. First switch to HQ may briefly stall due to larger bundle load + compile + memory setup.
• Steady-state speed is usually similar. For normal local-argmax decode, quality should match Lite.
• Use HQ when you need HQ-only APIs. Otherwise Lite is preferred for faster startup and smaller deploy size.

Note: any tiny transient slowdown right after switching bundles is expected; this is usually a one-time warm-up effect.