Complete zh-CN translation: add 75 missing strings across 6 files

[BeoKa-cita]

Summary

Complete the Chinese (Simplified) translation by adding 75 missing strings across 6 files:

• chat.json: Clear last used model, speculative decoding status, MCP plugin auth
• config.json: Max concurrent predictions, unified KV cache, chat template, MoE CPU offload layers, preset save changes dialog
• developer.json: API endpoint labels (Anthropic-compatible / OpenAI-like), Remote MCP allow/deny, server start/stop
• models.json: Memory guardrail warnings, GPU selection, model variant switching, "Indexing models..." status
• settings.json: Navigation bar position (top/left), Default Context Length settings, JIT TTL subtitle, Model Defaults label
• shared.json: Beta label

Checklist

• All translations follow existing zh-CN style conventions
• Template variables (e.g. {{count}}, <altOptionKey />) preserved
• Placeholder tokens ({{percentage}}, {{value}}) kept intact
• JSON syntax validated — all files parse correctly
• No keys removed — only additions and trailing whitespace cleanup

Co-Authored-By: Claude Opus 4.7 noreply@anthropic.com

[github-actions]

Thank you for your submission, we really appreciate it. Like many open-source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution. You can sign the CLA by just posting a Pull Request Comment same as the below format.

---

I have read the CLA Document and I hereby sign the CLA

---

_{You can retrigger this bot by commenting recheck in this Pull Request. Posted by the CLA Assistant Lite bot.}