Add zero-denominator guards to 8 trend indicators

[cinar]

Summary

Follow-up to #496 ("Tighten high-risk indicators to helper.Float"), which fixed the integer-truncation type-safety gap for these types but deliberately left their live float64 zero-denominator exposure unaddressed. This PR adds that guard to the 8 trend/ types from #496's follow-up list, each with a per-indicator fallback decided from its own documented formula/scale rather than an arbitrary value.

• Bop ((Close-Open)/(High-Low)): fallback 0. A zero-range bar (High == Low) forces Open == Close == High == Low too, so the numerator is also 0 in that case — 0 is the mathematically correct limit, not a guess, and matches BOP's signed, zero-centered scale (0 = equilibrium).
• Kama (Efficiency Ratio = Direction / Volatility): fallback 0. A perfectly flat window makes Volatility (moving sum of Abs(change)) zero, which also forces Direction to zero — ER = 0 ("no efficient movement occurred") matches most published KAMA implementations' explicit zero-volatility case.
• Kdj (RSV = (Close-Min)/(Max-Min)*100, then K/D/J derived from RSV): fallback 50. RSV is a 0-100 range ratio identical in shape to Stochastic's %K, so a zero range is treated as the neutral midpoint of that scale, same convention as RSI's existing flat-market fix (Fix RSI producing NaN instead of neutral 50 on flat market #455).
• Tsi (TSI = (PCDS/APCDS)*100): fallback 0. APCDS (smoothed absolute price change) is zero only when price has been perfectly flat — no momentum to report, which is TSI's neutral center on its signed -100..100 scale.
• Vwma (Sum(Price*Volume)/Sum(Volume)): fallback 0, documented as a known limitation. A window with zero volume (no trading) has no real volume-weighted price to report. Forward-filling the last valid average would arguably be more representative for plotting purposes, but requires per-window state that no other guard in this indicator family needs; 0 keeps the fix consistent with the other 7 stateless per-bar guards, and the limitation is called out explicitly in the doc comment.
• Cfo (((Price-Forecast)/Price)*100): fallback 0. Lowest-priority/weakest real-world risk (only triggers on a degenerate zero-price input) — 0 means "no forecast deviation to report."
• SlowStochastic (%K/%D, 0-100 range ratio): fallback 50, same neutral-midpoint reasoning as Kdj/RSI.
• Stochastic (%K/%D, 0-100 range ratio): fallback 50, same neutral-midpoint reasoning as Kdj/RSI.

Each guard is a plain if denom == 0 { return fallback } check immediately before the division, matching the existing pattern in momentum/ibs.go and volatility/chop.go. Doc comments on each type explain the chosen fallback and why.

Out of scope (per the task): momentum/volatility/volume package guards are separate parallel work; examples/, strategy/, README, and legal docs are untouched.

Test plan

• Added one flat/zero-range test per type (TestBopFlatBar, TestKamaFlatMarket, TestKdjFlatMarket, TestTsiFlatMarket, TestVwmaNoVolume, TestCfoZeroPrice, TestSlowStochasticFlatMarket, TestStochasticFlatMarket), each asserting the documented fallback value instead of NaN/Inf/panic.
• While writing TestKdjFlatMarket/TestSlowStochasticFlatMarket/TestStochasticFlatMarket I hit a real deadlock: these types fan out K/D(/J) internally through unbuffered DuplicateWithContext, which requires every output channel to be drained concurrently — fully draining one via helper.ChanToSlice before touching the others stalls the whole pipeline. Fixed by collecting each output channel on its own goroutine (sync.WaitGroup), matching the concurrent-read pattern the existing TestKdj/TestStochastic fixture tests already use via helper.CheckEquals.
• Verified all 8 existing CSV-fixture-based tests (TestBop, TestKama, TestKdj, TestTsi, TestVwma, TestCfo, TestSlowStochastic, TestStochastic) are unaffected — none of the fixtures contain a flat/zero-range bar, and all still pass byte-for-byte against their expected output.
• go build ./...
• go vet ./...
• gofmt -l . clean on all touched files (pre-existing unrelated gofmt drift elsewhere in the repo, outside this change's scope)
• go test ./... — all packages pass

https://claude.ai/code/session_01Xv4stuAb6WuQ8rPZ4cupLp

[codecov-commenter]

Codecov Report

❌ Patch coverage is 97.72727% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 92.01%. Comparing base (788025d) to head (3f1f939).

Files with missing lines	Patch %	Lines
trend/cfo.go	80.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #502      +/-   ##
==========================================
+ Coverage   91.96%   92.01%   +0.04%     
==========================================
  Files         232      232              
  Lines        7653     7661       +8     
==========================================
+ Hits         7038     7049      +11     
+ Misses        527      524       -3     
  Partials       88       88              

☔ 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.