Add Simplified Chinese README (collapsible section in README.md + standalone README_zh.md); English text unchanged

v2: re-quantized W4A16 (group 64, 200 iters, 234 real retrieval calibration texts); modest gains in BF16 fidelity and BEIR; README re-measured, build manifest added. v1 stays in git history.

docs: align Safetensors README with GGUF HQ tiers and document distinct quantization provenance

Update README: refine runtime notes, clarify size units, add companion GGUF link

Release Sakura-EmbeddingGemma-2-AutoRound W4A16

initial commit