feat: implement dual-pipeline signal engine service

New service at services/signal_engine/ implementing concurrent heuristic
(deterministic scoring) and probabilistic (Bayesian inference) pipelines
that evaluate technical signals across 6 timeframes (M30-M) and produce
independent BUY/WATCH/SKIP verdicts per ticker per evaluation tick.

Components:
- Input Normalizer: multi-source data assembly with sentinel fallbacks
- Signal Library: Fibonacci, MA Stack, RSI, Cup & Handle, Elliott Wave
- Multi-Timeframe Confluence Engine: weighted scoring with D/W/M anchors
- Hard Filter Engine: macro_bias, valuation, earnings proximity gating
- Heuristic Pipeline: S_total scoring with confidence-gated verdicts
- Probabilistic Pipeline: Bayesian log-odds with regime priors, entropy
  gating, EV_R calculation, and signal correlation penalty
- Exit Engine: stop-loss, targets, trailing ATR-based stops
- Delta Analyzer: pipeline agreement tracking with rolling Redis metrics
- Output Formatter: SignalOutput contract + Recommendation schema mapping
- Worker orchestrator: concurrent pipelines with failure isolation
- Main entry point: queue polling with fail-safe config loading

Infrastructure:
- Migration 039: signal_engine_outputs table with 3 indexes
- Helm chart: signalEngine service entry (processing tier)
- Redis key: QUEUE_SIGNAL_ENGINE constant

Tests: 390 tests (unit + property-based) covering all components
Config: dual_pipeline_enabled=false by default (safe rollout)

services/signal_engine/signals/fibonacci.py

@@ -0,0 +1,127 @@
+"""Fibonacci retracement signal evaluator.
+
+Computes retracement levels using ``L(r) = SH - r * (SH - SL)`` for the
+standard ratios [0.236, 0.382, 0.5, 0.618, 0.786] and produces a signal
+based on the proximity of the current price to the nearest level.
+"""
+
+from __future__ import annotations
+
+from services.signal_engine.models import OHLCVBar, SignalDirection, SignalResult
+from services.signal_engine.signals.base import (
+    find_swing_high,
+    find_swing_low,
+    validate_lookback,
+)
+
+# Standard Fibonacci retracement ratios
+RETRACEMENT_RATIOS: list[float] = [0.236, 0.382, 0.5, 0.618, 0.786]
+
+# Ratios considered "key" levels — proximity to these yields higher confidence
+_KEY_RATIOS: set[float] = {0.5, 0.618}
+
+# Default minimum number of bars required for evaluation
+DEFAULT_MIN_BARS: int = 20
+
+
+class FibonacciEvaluator:
+    """Fibonacci retracement signal evaluator.
+
+    Satisfies the :class:`~services.signal_engine.signals.base.SignalEvaluator`
+    protocol.
+
+    Parameters
+    ----------
+    min_bars:
+        Minimum number of OHLCV bars required before the evaluator will
+        produce a signal.  Defaults to ``20``.
+    """
+
+    def __init__(self, min_bars: int = DEFAULT_MIN_BARS) -> None:
+        self.min_bars = min_bars
+
+    # ------------------------------------------------------------------
+    # Public API (SignalEvaluator protocol)
+    # ------------------------------------------------------------------
+
+    def evaluate(
+        self,
+        bars: list[OHLCVBar],
+        timeframe: str,
+    ) -> SignalResult | None:
+        """Evaluate Fibonacci retracement on *bars* for *timeframe*.
+
+        Returns ``None`` when there are fewer than :pyattr:`min_bars` bars,
+        or when the swing high equals the swing low (flat market — no valid
+        retracement).
+        """
+        if not validate_lookback(bars, self.min_bars):
+            return None
+
+        # Detect swing high / swing low within the evaluation window
+        sh_result = find_swing_high(bars, self.min_bars)
+        sl_result = find_swing_low(bars, self.min_bars)
+
+        if sh_result is None or sl_result is None:
+            return None
+
+        _sh_idx, sh_price = sh_result
+        _sl_idx, sl_price = sl_result
+
+        # SH must be strictly greater than SL for a valid retracement range
+        if sh_price <= sl_price:
+            return None
+
+        price_range = sh_price - sl_price
+        current_price = bars[-1].close
+
+        # Compute retracement levels: L(r) = SH - r * (SH - SL)
+        levels: dict[float, float] = {
+            r: sh_price - r * price_range for r in RETRACEMENT_RATIOS
+        }
+
+        # Find the nearest retracement level to the current price
+        nearest_ratio: float = RETRACEMENT_RATIOS[0]
+        nearest_level: float = levels[nearest_ratio]
+        min_distance: float = abs(current_price - nearest_level)
+
+        for ratio in RETRACEMENT_RATIOS[1:]:
+            distance = abs(current_price - levels[ratio])
+            if distance < min_distance:
+                min_distance = distance
+                nearest_ratio = ratio
+                nearest_level = levels[ratio]
+
+        # Signal strength: 1.0 - (distance / range), clamped to [0, 1]
+        raw_strength = 1.0 - (min_distance / price_range)
+        strength = max(0.0, min(1.0, raw_strength))
+
+        # Direction: BULLISH if price is near a retracement level and above SL
+        # (potential bounce off support).  Otherwise BEARISH.
+        if current_price >= sl_price:
+            direction = SignalDirection.BULLISH
+        else:
+            direction = SignalDirection.BEARISH
+
+        # Confidence: higher when the nearest level is a key ratio (0.618, 0.5)
+        if nearest_ratio in _KEY_RATIOS:
+            confidence = min(1.0, strength * 1.2)
+        else:
+            confidence = strength * 0.8
+
+        return SignalResult(
+            signal_type="fibonacci",
+            timeframe=timeframe,
+            strength=strength,
+            direction=direction,
+            confidence=confidence,
+            metadata={
+                "swing_high": sh_price,
+                "swing_low": sl_price,
+                "retracement_levels": levels,
+                "nearest_ratio": nearest_ratio,
+                "nearest_level": nearest_level,
+                "distance_to_nearest": min_distance,
+                "current_price": current_price,
+            },
+        )