docs: restructure README/REPORT and clean metadata

Add outputs/benchmark.json prompts + images

Add examples/*.png (5x 1024 INT4 samples)

Add inference_int4.py

Add generate5.py

Add quantize_int4.py

Add REPORT.md

Add README with examples + benchmark table

Upload Qwen-Image-2.1 OpenVINO INT4 (NNCF weight-only, transformer+text_encoder 4bit)

initial commit