Improve RMA indicator parity with Cython

[nicolad]

Improve WilderMovingAverage (RMA) indicator parity with Cython #2507

Restores feature-parity and behavioural equivalence between the Rust WilderMovingAverage implementation and the canonical Python/Cython reference.

---

1 · Why this matters — Context & Motivation

Theme	Why it matters
Correctness ⇄ Consistency	Mixed Python (quant research) + Rust/WASM (execution) stacks must emit identical indicator values; silent divergence leaks P n L.
Predictable warm-ups & resets	Edge-cases (e.g. period = 1, reset() after NaN) previously left the indicator in an undefined state.
Baseline for SIMD hot-paths	Before vectorising we need a byte-for-byte port; this PR closes the last functional gaps.
Safety-first API	Panics with descriptive messages before UB occurs, mirroring Python checks.
Test-driven culture	Full parity test-suite protects against future regressions and enables confident refactors.

---

2 · What changed

Concern	Python/Cython (🏆 truth)	Rust before	Rust after (this PR)
Parameter validation	✔ period > 0	✘ none	✔ assert!(period > 0) with descriptive panic
alpha computation	✔ 1 / period	✔ same formula but duplicate field	✔ single source-of-truth, removes redundant assignment
price_type propagation	✔ all inner handlers inherit	✘ mixed defaults	✔ explicit PT cascades to handle_* methods
First-sample seeding	✔ seeds & count = 1	✘ count = 0	✔ aligned
count increments	✔ on every update	✘ skipped first	✔ parity
period = 1 fast-path	✔ RMA ≡ last price	✘ undefined	✔ unit-tested degeneracy
reset() semantics	✔ full zero-state	✘ stale has_inputs	✔ cleared
Display::fmt	n/a	✘ trailing comma	✔ fixed
has_inputs lifecycle	n/a	✘ false-positive after reset()	✔ toggled only on first sample, cleared on reset()
Thread-safety docs	n/a	✘ missing	✔ explicit not Send + Sync
Test coverage	full	partial	full parity + 🆕 rstest suite (18 cases, 100 % line cov)

---

3 · Five practical ways to put RMA to work

#	Use-case	Why RMA helps	Wiring sketch
1	ATR-style volatility filter	Wilder’s MA is the core of ATR; re-use it directly for smoother σ estimates.	atr = true_range.rma(14)
2	Mean-reversion bands	Slower reaction than EMA; draws bands that fade whipsaws.	long when price < RMA - k·σ
3	Stop-loss decay	α = 1/period gives exponential decay ideal for trailing stops.	stop = max(stop, RMA(20))
4	Liquidity / impact model	RMA of volume quantifies structural flow better than SMA.	liq = volume.rma(30)
5	Momentum factor research	Wilder MA is default in many CTA specs; parity allows backtest ⇄ live reuse.	nightly batch: rank = close/RMA(100) - 1

---

4 · Implementation notes

• Guardrails — period > 0 panic mirrors Python; prevents div-by-zero UB.
• period = 1 degeneracy — α = 1 → RMA = last sample; test shields future optimisers.
• update_raw — inlined FMA (mul_add) for numerical precision.
• has_inputs — explicit, avoids needless Option<f64> allocations.
• NaN poisoning — once a NaN enters, value remains NaN until reset() (same as reference).
• Thread-safety — indicator is not Send + Sync; wrap in Arc<Mutex<>> for multi-thread.
• Public API — no breaking signature changes; only stricter panics.
• Performance heads-up (speculative) — with parity achieved, next step is to vectorise the hot-path (update_raw) using nightly core::simd. 🚧 Flagged as prediction.

---

5 · Tests added / updated

Test name	Purpose
test_new_with_zero_period_panics	Validate guardrail panic message
first_tick_seeding_parity	Ensure first sample seeds value & count = 1
numeric_parity_with_reference_series	Byte-for-byte parity vs Python reference on canonical series
test_rma_period_one_behaviour	Period = 1 degeneracy correctness
test_rma_large_period_not_initialized	Warm-up exactness on large periods
test_reset_reseeds_properly	Reset restores zero-state
test_default_price_type_is_last	Default PriceType::Last sanity check
test_update_with_nan_propagates	NaN propagation rules
+ 10 additional parity & robustness cases	18 total, 100 % line/branch coverage

---

• Related Confirm parity between Python and Rust indicators #2507

---

“The trend is your friend… until it ends.”
— J. Welles Wilder Jr.

[cjdsellers]

@nicolad excellent work and thanks for all the tests! 👌