Switch to ONNX runtime with chatterbox-turbo-ONNX (fp16)

Replaces the PyTorch/chatterbox-tts stack with direct ONNX inference
using ResembleAI/chatterbox-turbo-ONNX fp16 weights.

- engine.py: full rewrite — ONNX sessions, autoregressive KV-cache LM
  loop, voice conditionals cache via speech_encoder outputs
- wyoming_handler.py: remove torch dep, use np.asarray for audio bytes
- requirements-rocm-init.txt: onnxruntime-rocm replaces torch wheels
- requirements-rocm.txt: drop chatterbox/torch deps, keep audio utils
- Dockerfile.rocm: remove chatterbox-tts install step

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Dockerfile.rocm

@@ -17,18 +17,15 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
 
 WORKDIR /app
 
-# Step 1: Install ROCm-compatible PyTorch stack first.
-# This must happen before anything else to prevent pip from pulling CPU wheels.
+# Step 1: Install onnxruntime-rocm first so it claims the onnxruntime namespace
+# before any other package can pull in the CPU-only onnxruntime wheel.
 COPY requirements-rocm-init.txt .
 RUN pip3 install -r requirements-rocm-init.txt
 
-# Step 2: Install remaining dependencies (pinned to avoid overwriting torch).
+# Step 2: Install remaining dependencies.
 COPY requirements-rocm.txt .
 RUN pip3 install -r requirements-rocm.txt
 
-# Step 3: Install chatterbox with --no-deps so pip cannot replace ROCm torch.
-RUN pip3 install --no-deps chatterbox-tts
-
 # Application source
 COPY engine.py config.py wyoming_handler.py wyoming_voices.py main.py ./