feat: add ADX filter to block sideways market entries

ADX < 25 now returns HOLD in get_signal(), preventing entries during
trendless (sideways) markets. NaN ADX values fall through to existing
weighted signal logic. Also syncs the vectorized dataset builder with
the same ADX filter to keep training data consistent.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

src/dataset_builder.py

@@ -47,6 +47,10 @@ def _calc_indicators(df: pd.DataFrame) -> pd.DataFrame:
     d["stoch_k"] = stoch["STOCHRSIk_14_14_3_3"]
     d["stoch_d"] = stoch["STOCHRSId_14_14_3_3"]
 
+    # ADX (14) — 횡보장 필터
+    adx_df = ta.adx(high, low, close, length=14)
+    d["adx"] = adx_df["ADX_14"]
+
     return d
 
 
@@ -112,6 +116,12 @@ def _calc_signals(d: pd.DataFrame) -> np.ndarray:
     # 둘 다 해당하면 HOLD (충돌 방지)
     signal_arr[long_enter & short_enter] = "HOLD"
 
+    # ADX 횡보장 필터: ADX < 25이면 추세 부재로 판단하여 진입 차단
+    if "adx" in d.columns:
+        adx = d["adx"].values
+        low_adx = (~np.isnan(adx)) & (adx < 25)
+        signal_arr[low_adx] = "HOLD"
+
     return signal_arr