Add NVIDIA Nemotron 3 Super deployment guide example page

[wbrennan899]

Summary

• Add end-to-end guide for deploying NVIDIA Nemotron 3 Super (120B/12B active MoE) on Vast.ai using SGLang
• Covers instance search, deployment with FP8 quantization, and querying via OpenAI-compatible API
• Documents all three reasoning modes (on, off, low-effort) with Python and cURL examples
• Includes note about SGLang's nano_v3 parser behavior where reasoning-off responses are returned in reasoning_content instead of content