Add quant risk and sizing research features

[AlanBuildsAI]

Summary

This branch upgrades NeuroQuantAI with a more quant-serious research layer while keeping the project offline-first, honest, and safe.

It adds:

• structured fee / spread / slippage cost modeling
• position sizing methods
• optional risk controls
• gross vs net return tracking
• exposure diagnostics
• benchmark support through benchmark_close
• rolling risk diagnostics
• overfit gap detection
• transparent robustness score
• synthetic sample CSV mode
• additional Streamlit controls and output tabs
• expanded tests

Safety / scope

This does not add:

• live trading
• broker/API connections
• API keys
• order routing
• OAuth
• real broker integrations
• live market data calls

The bundled sample CSV is fully synthetic and included only to demonstrate the expected CSV schema.

Verification

• Pipeline runs cleanly
• Streamlit compiles
• Streamlit boots locally with HTTP 200 / health OK
• Test suite passes: 112 tests

Notes

This PR should be reviewed before merging because main is already deployed and featured publicly.