feat: Phase 2b — data core (DataPoint, Indicators, Levels, SimulationData)

[may01]

Summary

• Add DataPoint protocol + LiveDataPoint / WideDataPoint — unified read interface so signals/strategies are path-agnostic (live vs simulation)
• Add Indicators class with TA-Lib path and pandas fallback; all columns named {tf}_{indicator} and computed in-place
• Add Levels class for support/resistance price level management
• Add SimulationData — O(1) backtesting replay engine over pre-computed wide DataFrame

Depends On

PR #1 (Phase 2a — BaseStock, BinanceStock, StockItem)

Test Plan

• pytest tests/data/ — 38 passed, 0 failed
• LiveDataPoint.get(col, tf, shift) returns correct scalar; raises KeyError for missing tf
• WideDataPoint.get() uses position-based indexing with bounds guard
• Indicators pandas fallback: RSI in [0,100], EMA columns present, idempotent, drop_na shrinks df
• Indicators index-aligned (works with datetime-indexed DataFrames)
• Levels: add/load/nearest/clear all correct; load deep-copies dicts
• SimulationData: step advances cursor, is_done, reset, StopIteration on exhaustion

🤖 Generated with Claude Code