Add vLLM NVFP4 lm_head patch (ParallelLMHead -> ReplicatedLinear)

Add model card (all-FP4 max-speed variant; quality note; serve + lm_head patch)

Add preprocessor_config.json for vision support

Add ParallelLMHead to quantization targets for vLLM compat

NVFP4-only quant v2 (all Linear incl MTP → NVFP4)

initial commit