docs: enhance README, add comprehensive documentation and Claude skill

[Claude]

Added comprehensive documentation and Claude skill to teach AI agents how to use the file-based trading system.

Changes

Enhanced README

• Added badges and expanded introduction with core philosophy
• Added use cases section (AI trading, backtesting, audit trails)
• Added advanced features section covering:
  • Batch order processing
  • Order status tracking (EXECUTION/REJECTION)
  • Ledger archiving and compression
  • Market data formats (JSON + text)
• Added troubleshooting guide
• Added security recommendations and FAQ

Created Documentation (docs/)

• architecture.md: System design, component diagrams, data flows, extension points
• ai-agent-guide.md: Complete Python examples for AI agents including order submission, quote fetching, P&L tracking, risk control
• order-format.md: Beancount format specification with field requirements, formatting rules, common errors
• README.md: Documentation index

Created Claude Skill (.claude/skills/longbridge-trading/)

Comprehensive skill teaching AI how to:

• Submit buy/sell orders by appending to fs/trade/beancount.txt
• Request quotes by creating fs/quote/track/{SYMBOL} files
• Check status by reading JSON files (state.json, pnl.json, portfolio.json)
• Configure risk control via risk_control.json
• Includes prerequisites, workflows, error handling, symbol formats

Example: AI Agent Usage

# Submit a limit order
cat >> fs/trade/beancount.txt << 'EOF'
2026-02-12 * "ORDER" "BUY AAPL.US"
  ; intent_id: 20260212-001
  ; side: BUY
  ; symbol: AAPL.US
  ; qty: 100
  ; type: LIMIT
  ; price: 180.50
  ; tif: DAY

EOF

# Wait for controller to process (2s)
sleep 3

# Check result
tail -20 fs/trade/beancount.txt

All documentation is bilingual (Chinese with English code examples) and optimized for both human developers and AI agents.