Optimizer Results Schema and Grid Ranking

[Muratich]

Depends on: Samy GridOptimizer issue, #111
Design doc: docs/strategy_composable_engine_design.md (§1.7)
Docs: docs/analytics_data_model_specification.md

Two ranking types — do not mix them:

1. Strategy ranking (already exists): build_top_n() ranks different strategies
   in the dashboard catalogue. Uses TopNEntry (strategy_id, instrument, metrics).

2. Optimizer ranking (this issue): rank param combos for ONE strategy after
   a grid search. Each row has rank + params + metrics.

Scope:

• Define optimizer output JSON schema:
  { strategy_id, instrument, ranked: [{ rank, params, metrics }] }
• Add rank_optimizer_results(reports, config) -> list[OptimizerRankedEntry]
  Input: list of (params dict, MetricsReport) from GridOptimizer runs
• Reuse RankingConfig sort order (P&L → drawdown → Sharpe → win rate)
  Do NOT call build_top_n() for optimizer results
• Skip invalid / empty runs (non-finite metrics, empty trade log)
• Document schema in docs/analytics_data_model_specification.md
• Agree JSON field names with Samy before Muratich wires #115

Out of scope:

• GridOptimizer implementation (Samy)
• Dashboard UI (#123, Week 6)
• New metrics formulas
• Legacy Python strategy optimizer (Week 6 / P1)

Definition of Done:

• OptimizerRankedEntry (or equivalent) includes rank, params, metrics
• rank_optimizer_results() with unit test: 5 fake grid results → stable top-3
• Schema documented in analytics_data_model_specification.md
• Works with composable strategy_id (e.g. ma_rsi_composable)