Live IG execution, experiments/analysis subsystem, rolling-window queue ladder, new strategies and docs overhaul

[mccaffers]

This issue is addressed through the optimisation of the backtesting engine, with a research path, and by connecting the dots in a live trading execution.

Optimise Backtesting & Research Capabilities

• Add a rolling-window chaining to use a strict-priority queue ladder with performance gating, ensuring underperforming runs stop chaining and stay out of winner indices.
• Support cross-run tick superset cache.
• Add new strategies (NY open-range breakout, liquidity-sweep reversal, range-bar velocity) to bring our total to nine, and integrate a shared time-cap exit across the entire book.
• Build out an experiments/analysis subcommand pair to answer conditional price behaviour questions and log MFE/MAE quantiles, time buckets, and spread stats directly to ElasticSearch. The current implementation focuses on trades; I need a pre-trading environment to experiment with as well.

---

Live Trading

• Add a IG order path featuring confirm polling, rate-budget tracking, duplicate suppression, trade locks, cluster-exposure gating, and a fail-closed-opens / fail-open-closes safety policy.
• Support the deployment of an in-engine positions producer to mirror the broker book to Redis every minute, along with a tracking subcommand to enrich the deal feed into our position book and audit log.
• Enforce winner selection gating based on score, max drawdown, and Calmar ratio over full terminal histories.