Consolidate zero-denominator guards into helper.SafeDivide

[cinar]

Summary

Earlier work independently added zero-denominator/NaN guards across ~17 indicators in trend, momentum, volatility, and volume, each shaped like:

if denom == 0 {
    return fallback // 0, 50, or 0.5
}
return numerator / denom

This PR adds a single generic helper, helper.SafeDivide[T Number](numerator, denominator, fallback T) T (in helper/safe_divide.go, modeled on the existing scalar helper helper.RoundDigit), and refactors every matching call site to use it. Behavior is unchanged — same fallback values, same floating-point evaluation order — verified by the full existing test suite passing unmodified.

Call sites refactored

Direct pattern (SafeDivide(num, denom, fallback) replaces the guard 1:1):

• trend/kama.go — Efficiency Ratio, fallback 0
• trend/bop.go — BOP, fallback 0
• trend/vwma.go — VWMA, fallback 0
• momentum/ibs.go — Internal Bar Strength, fallback 0
• momentum/ultimate_oscillator.go — per-period average, fallback 0.5
• volatility/bollinger_band_width.go — fallback 0
• volatility/z_score.go — fallback 0
• volatility/acceleration_bands.go — fallback 0
• volatility/percent_b.go — %B, fallback 0.5
• volume/mfm.go — Money Flow Multiplier, fallback 0
• volume/cmf.go — CMF, fallback 0

Scaled pattern (original code applied a * 100 or pre-multiplied the numerator by 100 around the guarded division; to keep the arithmetic byte-for-byte identical, the fallback is expressed in the same pre-scale units so the surrounding multiply reproduces the exact original float value):

• trend/kdj.go — RSV: SafeDivide(num, denom, 0.5) * 100 (0.5 × 100 = 50, exact)
• trend/slow_stochastic.go — fastK: same transform
• trend/stochastic.go — %K: same transform
• momentum/stochastic_oscillator.go — fastK: same transform
• trend/tsi.go — SafeDivide(pcd, apcd, 0) * 100 (fallback 0, scaling is a no-op)
• trend/cfo.go — SafeDivide(num, price, 0) * 100 (same)
• volatility/po.go — original pre-multiplied the numerator (100 * (closing - pl) / denom), so here it's SafeDivide(100*(closing-pl), denom, 50) — fallback needs no rescaling since the original guard already returned the unscaled 50 literal

helper.SafeDivide is constrained to helper.Number (not helper.Float) specifically so that trend/tsi.go, whose Tsi[T helper.Number] is generic over integers too, can call it directly without widening or narrowing any existing type's constraint.

Explicitly skipped (do not fit the pattern)

• volume/vwap.go — carries forward the last valid computed value instead of a constant fallback (per task instructions, left untouched).
• trend/mcginley_dynamic.go — same shape as Vwap: on a zero previous value it reseeds state and takes a completely different code path (not a single guarded return of a quotient), so it isn't a SafeDivide candidate.
• momentum/rsi.go — the guard condition is averageGain == 0 && averageLoss == 0 (compound, two variables), not a single denominator check, so it doesn't fit SafeDivide(numerator, denominator, fallback)'s shape.
• volatility/chop.go — the guarded division feeds into math.Log10(sum/diff) rather than being returned directly as a quotient; the guard and the formula shape don't reduce to a plain numerator/denominator.

Test plan

• go build ./...
• go vet ./...
• gofmt -l . (no new files flagged; pre-existing unrelated formatting debt in unrelated examples//strategy//backtest test files, untouched by this PR)
• go test ./... -timeout 300s — all packages pass unmodified, confirming no behavior change

🤖 Generated with Claude Code

https://claude.ai/code/session_01Xv4stuAb6WuQ8rPZ4cupLp

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 91.99%. Comparing base (74b9529) to head (f998834).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #505      +/-   ##
==========================================
+ Coverage   91.96%   91.99%   +0.02%     
==========================================
  Files         233      234       +1     
  Lines        7719     7703      -16     
==========================================
- Hits         7099     7086      -13     
+ Misses        531      528       -3     
  Partials       89       89              

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.