docs: update benchmark results for LoCoMo and RAG

[iAlexKai]

Update benchmark results in README files:

Changes:

• Update LoCoMo benchmark results in README_CN.md (Table 1 data with gpt-4o-mini judge)
• Add new section '1.2 Key Efficiency Improvements' in README.md
• Add new section '3.2 Single-turn RAG' in README.md
• Synchronize updates to README_JA.md
• Remove redundant benchmark paragraphs and OpenClaw Context Plugin Details section
• Fix Chinese documentation links to point to docs/zh/ directory
• Remove 'Academic Backing' / '学术背书' / '学術的裏付け' sections

Files modified:

• README_CN.md
• README.md
• README_JA.md

[github-actions]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

⏱️ Estimated effort to review: 2 🔵🔵⚪⚪⚪

🏅 Score: 95

🧪 No relevant tests

🔒 No security concerns identified

✅ No TODO sections

🔀 No multiple PR themes

⚡ No major issues detected

[github-actions]

PR Code Suggestions ✨

No code suggestions found for the PR.

[ZaynJarvis]

lgtm