fix(code-review): correct README algorithm description and add tests/lint.sh

[kpatel513]

Summary

The README documented a confidence-scoring algorithm (0–100, threshold 80) that was replaced by a validate/filter subagent approach. It also described Agent 4 as a "git blame/history analyzer" when it actually scans introduced code for security issues and incorrect logic. Anyone reading the README to understand how the plugin works was reading a description of a previous implementation.

Changes

plugins/code-review/README.md

• Replace all confidence-scoring language with accurate validate/filter description
• Fix Agent 4 description: "git blame/history analysis" → "security and logic review of introduced code"
• Remove "Confidence scoring" section (0/25/50/75/100 scale that doesn't exist)
• Remove "Adjusting confidence threshold" configuration section (referenced code that doesn't exist)
• Replace "Scoring system" technical details with accurate "Validation system" description

plugins/code-review/tests/lint.sh

• Checks allowed-tools and description frontmatter fields are present
• Checks any references/*.md files cited in command files exist on disk
• Stateless, fast, exits non-zero on failure

Test plan

• Run bash plugins/code-review/tests/lint.sh — should pass 2/2
• Read through README and verify it accurately matches commands/code-review.md