perf: Parallelize signal generation strategies for faster execution

Summary

The signal generation script (scripts/signals/signals_all.py) currently runs 10 trading strategies sequentially, despite them being functionally independent. This results in long execution times (often 2-3+ minutes total).

Current Behavior

• 10 strategies execute one after another
• Each strategy loads its own market data independently
• No data sharing or caching between strategies
• Console output interleaved with execution
• Typical strategy times: 12-25 seconds each

Example from recent run:

MOC_LONG: 23.4s
MOC_SHORT: 18.1s
MR_LONG: 12.2s
TREND: 18.4s
... (total ~2-3 minutes)

Desired Behavior

• Run strategies in parallel to reduce total execution time
• Maintain orderly console output (strategies print in consistent order)
• Maintain orderly log file output for debugging/verification
• Handle errors gracefully without affecting other strategies

Technical Considerations

1. Strategies are independent - Each receives read-only current_positions_df and usable_capital
2. Norgate API - May have rate limits on concurrent requests
3. Output ordering - Need to buffer/queue output to maintain readability
4. Logging - Python's logging module is thread-safe but output interleaving is a concern

Proposed Options

1. Multi-processing with separate modules - Each strategy in its own file, run as subprocess
2. concurrent.futures ThreadPoolExecutor - Simpler but GIL-limited
3. concurrent.futures ProcessPoolExecutor - True parallelism within single script
4. Hybrid approach - Group strategies by market (US vs AU) and run groups in parallel

Requirements

• [ ] Orderly console output (print strategies in consistent order after completion)
• [ ] Orderly log output for verification
• [ ] Error isolation (one strategy failure doesn't stop others)
• [ ] Configurable parallelism (--parallel flag or config setting)

🤖 Generated with Claude Code