Fix: Correct base model from Qwen3-2B to Qwen3-8B

Update to continuous batching model (batch_size=4, neuronxcc 2.21)

Add pipeline_tag metadata and model documentation

Upload Neuron-traced ChessLM model for vLLM inference on AWS Trainium

initial commit