Add first-class backtest engine with standard metrics

[TDamiao]

Goal

Provide a first-class backtest module to turn indicators/signals into decisioning results with standardized portfolio metrics.

Why

The library already has strong indicator + streaming + compatibility foundations. A built-in backtest path is the biggest adoption multiplier for strategy evaluation.

Implementation prompt

Implement backtest/ primitives that run deterministic historical simulations over OHLCV bars:

• Inputs: candles/returns + signals or strategy callback
• Position model: flat/long/short with configurable fees/slippage
• Execution model: close-to-close and next-bar execution options
• Output: equity curve, trades, and metrics object

Implement metric helpers:

• Sharpe
• Sortino
• Max Drawdown
• Profit Factor
• Win Rate
• Expectancy

Scope

• New public exports under root and ta-crypto/backtest
• Typed contracts for trades, equity curve, and summary metrics
• Deterministic results for same inputs/config
• Clear warmup handling and NA-safe calculations

Acceptance criteria

• runBacktest(...) produces reproducible trade list + equity curve
• Metric package returns stable values for known fixtures
• README includes one end-to-end example from signal -> report
• Unit tests cover fees/slippage, no-trade paths, and drawdown edge cases

Test plan

• Golden fixture for one known strategy path
• Regression tests for metrics with hand-checked expected values
• Cross-check one scenario against an external reference notebook/script

[TDamiao]

Tracked by v0.4 master gate: #21.