Add slide and rotate interactive captcha solvers

New solver subsystem with independent models:
- GapDetectorCNN (1x128x256 grayscale → sigmoid) for slide gap detection
- RotationRegressor (3x128x128 RGB → sin/cos via tanh) for rotation angle prediction
- SlideSolver with 3-tier strategy: template match → edge detect → CNN fallback
- RotateSolver with ONNX sin/cos → atan2 inference
- Generators, training scripts, CLI commands, and slide track utility

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

training/dataset.py

@@ -224,3 +224,55 @@ class RegressionDataset(Dataset):
             img = self.transform(img)
         return img, torch.tensor([label], dtype=torch.float32)
 
+
+# ============================================================
+# 旋转求解器用数据集 (sin/cos 编码)
+# ============================================================
+class RotateSolverDataset(Dataset):
+    """
+    旋转求解器数据集。
+
+    从目录中读取 {angle}_{xxx}.png 文件，
+    将角度转换为 (sin θ, cos θ) 目标。
+    RGB 输入，不转灰度。
+    """
+
+    def __init__(
+        self,
+        dirs: list[str | Path],
+        transform: transforms.Compose | None = None,
+    ):
+        """
+        Args:
+            dirs:      数据目录列表
+            transform: 图片预处理/增强 (RGB)
+        """
+        import math
+
+        self.transform = transform
+        self.samples: list[tuple[str, float, float]] = []  # (路径, sin, cos)
+
+        for d in dirs:
+            d = Path(d)
+            if not d.exists():
+                continue
+            for f in sorted(d.glob("*.png")):
+                raw_label = f.stem.rsplit("_", 1)[0]
+                try:
+                    angle = float(raw_label)
+                except ValueError:
+                    continue
+                rad = math.radians(angle)
+                self.samples.append((str(f), math.sin(rad), math.cos(rad)))
+
+    def __len__(self) -> int:
+        return len(self.samples)
+
+    def __getitem__(self, idx: int):
+        import torch
+        path, sin_val, cos_val = self.samples[idx]
+        img = Image.open(path).convert("RGB")
+        if self.transform:
+            img = self.transform(img)
+        return img, torch.tensor([sin_val, cos_val], dtype=torch.float32)
+