fix(features): revert to 2-channel default; OHLCV features redundant

HPO finding: hl_range≈realized_vol, ret_intrabar≈ret — correlation kills signal.
4ch D=128: 0.3503, 4ch D=256: 0.3807, 2ch D=128 baseline: 0.3908 (winner).
Parquet keeps hl_range+ret_intrabar; comment in build() documents the attempt.
test_build_uses_4_channels → test_build_uses_2_channels (tracks current default).

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

train.py

@@ -154,10 +154,9 @@ def build():
     else:
         df = pd.read_parquet(daily_path).reset_index(drop=True)
         df["date"] = pd.to_datetime(df["date"])
-    # Use OHLCV-derived features when available; fall back to 2-channel
-    base_feats = ["ret", "realized_vol"]
-    extra_feats = [c for c in ["hl_range", "ret_intrabar"] if c in df.columns]
-    FEAT_COLS = base_feats + extra_feats
+    # 2-channel default (HPO: adding hl_range+ret_intrabar hurt — correlated with base feats)
+    # To experiment: change to ["ret", "realized_vol", "hl_range", "ret_intrabar"]
+    FEAT_COLS = ["ret", "realized_vol"]
     feats  = df[FEAT_COLS].to_numpy(np.float32)
     target = df["realized_vol"].to_numpy(np.float32)
     tr_idx = df.index[df["date"].dt.year <= 2021].tolist()