feat(family-15): add 17 risk/performance metrics

[kingchenc]

Summary

Adds Family 15 — Risk / Performance as 17 standard Indicators, taking
the codebase from 71 → 88 indicators across nine families. Implemented
pragmatically as Indicator<f64> / Indicator<(f64, f64)> rather than a
separate wickra-metrics crate (as the original
indicator-ideas/families/15-risk-performance.md suggested) — keeps the
public surface consistent and lets every binding pick the metrics up for
free.

14 scalar metrics (Indicator<Input = f64>)

Input = period return / equity sample / per-trade P&L.

• SharpeRatio(period, risk_free) — (mean − rf) / sample_stddev.
• SortinoRatio(period, mar) — Sharpe with downside-only volatility.
• CalmarRatio(period) — mean return / max drawdown of equity curve.
• OmegaRatio(period, threshold) — Σ gains-above-threshold / Σ losses-below.
• MaxDrawdown(period) — rolling peak-to-trough decline, monotone deque.
• AverageDrawdown(period) — mean drawdown depth (= Pain Index by another name).
• DrawdownDuration — cumulative bars under water, output u32.
• PainIndex(period) — Becker's mean-dd risk measure.
• ValueAtRisk(period, confidence) — historical VaR with linear-interpolated percentile.
• ConditionalValueAtRisk(period, confidence) — expected loss in tail (CVaR / ES).
• ProfitFactor(period) — Σ gains / Σ losses.
• GainLossRatio(period) — avg win / avg loss.
• RecoveryFactor — cumulative net return / max drawdown.
• KellyCriterion(period) — win_rate − (1 − win_rate) / payoff.

3 pair metrics (Indicator<Input = (f64, f64)>)

Input = (asset_return, benchmark_return).

• TreynorRatio(period, risk_free) — (mean_asset − rf) / Beta_vs_bench.
• InformationRatio(period) — mean(active) / tracking_error.
• Alpha(period, risk_free) — Jensen's Alpha (CAPM regression).

Touchpoints

• Core: 17 new files under crates/wickra-core/src/indicators/,
  mod.rs + crates/wickra-core/src/lib.rs re-exports.
• Python bindings: 17 new #[pyclass]es + add_class registrations
  • __init__.py __all__ extension.
• Node bindings: 17 new #[napi] wrappers (scalar + pair) +
  index.js destructure / exports.
• WASM bindings: 12 scalar metrics via the existing
  wasm_scalar_indicator! macro + 5 handwritten (DrawdownDuration,
  RecoveryFactor, Treynor, InformationRatio, Alpha).
• Fuzz: scalar metrics appended to fuzz_targets/indicator_update.rs,
  new fuzz_targets/indicator_update_pair.rs for (f64, f64) inputs
  (registered in fuzz/Cargo.toml).
• Python tests: SCALAR list extended in test_new_indicators.py,
  new PAIR parameter list with test_pair_streaming_matches_batch,
  17 reference-value cases in test_known_values.py.
• Node tests: scalar factories + new pair-factory block in
  bindings/node/__tests__/indicators.test.js.
• Benches: 5 new Family-15 benches in
  crates/wickra/benches/indicators.rs.
• Docs: README family-table row, indicator counter 71 → 88 (line 112
  • structure diagram), CHANGELOG.md [Unreleased] entry.
• Wiki drafts (not pushed; live in indicator-ideas/families/wiki/family-15-risk-performance/
  in this branch's working tree but ghost-ignored — 17 page drafts +
  _Sidebar.md / Indicators-Overview.md / Warmup-Periods.md / Home.md
  deltas).

Notes on Family 12 / merge conflict

Family 12 (PR #51 — feat/family-12-statistik-regression) introduces
node_pair_indicator! and wasm_pair_indicator! macros for the
Indicator<(f64, f64)> pattern (Pearson / Beta / Spearman). Family 12
is not yet merged into main and this PR also needs that pattern —
the three pair wrappers (Treynor / InformationRatio / Alpha) are
therefore written by hand here, plus a header note in
bindings/node/src/lib.rs and bindings/wasm/src/lib.rs flagging the
conflict.

Merge resolution when PR #51 lands: keep the macros from Family 12
and replace the three handwritten Family 15 pair wrappers with macro
invocations:

// node
node_pair_indicator!(TreynorRatioNode, "TreynorRatio", wc::TreynorRatio);    // needs risk_free arg
node_pair_indicator!(InformationRatioNode, "InformationRatio", wc::InformationRatio);
node_pair_indicator!(AlphaNode, "Alpha", wc::Alpha);                          // needs risk_free arg

Two of the three (Treynor / Alpha) have an extra risk_free parameter
that the existing node_pair_indicator! macro doesn't take — those will
need a small macro generalisation or stay handwritten. Trivial.

Stolpersteine encountered

1. MaxDrawdown semantics: implemented as rolling-window only
   (consistent with other rolling indicators). Setting period ≥ full
   bar count gives effectively cumulative behaviour. RecoveryFactor
   uses cumulative-from-start.
2. VaR / CVaR rolling: VecDeque + clone + sort per update,
   O(period log period). Adequate for typical period ≤ 252.
3. Edge cases: flat windows (zero stddev / variance / drawdown) return
   0.0 rather than NaN; no-loser windows for Profit Factor / GLR /
   Omega return f64::INFINITY; Kelly clamps to −1.0 on all-loser
   windows.
4. DrawdownDuration output is u32, not f64 — Node batch() and
   Python batch() widen to f64 (NaN for warmup, integer values cast
   to float) to keep Float64Array / np.float64 array contract.

Test plan

• cargo check --workspace --all-features (green).
• cargo test -p wickra-core --lib (740 tests pass, includes the
  new families).
• CI: 28 jobs + codecov/patch on the PR.
• Python pytest bindings/python post-build (deferred to CI).
• Node npm test post-build (deferred to CI).

Per the project workflow, full cargo test --workspace / cargo clippy
runs are left to CI; this PR only ran cargo check locally per the
lean-batch instruction.

[codecov]

Codecov Report

❌ Patch coverage is 99.78202% with 4 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
crates/wickra-core/src/indicators/calmar_ratio.rs	99.06%	1 Missing ⚠️
...ates/wickra-core/src/indicators/recovery_factor.rs	99.08%	1 Missing ⚠️
crates/wickra-core/src/indicators/treynor_ratio.rs	99.18%	1 Missing ⚠️
crates/wickra-core/src/indicators/value_at_risk.rs	99.09%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!