docs: quantized-from-Ornith (Option A) + gfx1201/radeon-ai-pro-r9700 tags (RDNA4 discoverability)

docs: quantized-from-Ornith (Option A) + gfx1201/radeon-ai-pro-r9700 tags (RDNA4 discoverability)

Add quantize_mxfp4.py (step-1 MXFP4 quant recipe) + link it in the card

card: drop untested experiential note; credit Rob Smith (tcclaviger) for the RDNA4 vLLM base

model tree: list both parents (DeepReinforce base + protoLabs MTP head), relation=merge

model card