feat(web): RWA market-making simulator dashboard

[abhicris]

Adds a zero-build single-page interactive simulator for the meridian agent.

What it shows

• Switch between constant_spread and adaptive_spread and watch quotes / fills / inventory / PnL respond live.
• Four asset presets (private credit, T-bill, real estate, long-tail ERC-20) — each sets a sane default vol regime.
• Sliders for base spread, order size, max inventory skew, inventory penalty, market vol, fill probability, toxic-flow share, tick speed.
• KPIs: realised PnL, inventory PnL, effective spread, fill rate, adverse-selection rate, sharpe-ish over a 30-tick window.
• Live quote ladder, recent-fills table.
• Strategy roadmap section: compliance_gated and geo_priced stubs that point at the dependency on kcolbchain/rwa-toolkit.

Design

• Single web/index.html, no build, no backend.
• Same dark palette + layout family as the other kcolbchain dashboards.
• Auto-runs a Brownian mid simulation so first-time visitors see motion.
• Hosts at kcolbchain.com/meridian/ once deployed.

Out of scope (Python codebase still owns)

• Live venue connectivity.
• Historical backtesting (src/backtest/engine.py).
• The richer Avellaneda-Stoikov / inventory-skew math — the in-browser adaptive_spread is a teaching simplification.

Test plan

• Open locally; quotes + ladder render within 1 tick.
• Adverse-selection KPI rises as toxic-flow slider moves up.
• Inventory PnL flips sign when inventory crosses zero.
• No console errors in modern Chrome/Safari.