Operator: HardSwish-Affine-Gate (Fused CUDA Kernel)

Definition
- z = x * scale + bias
- m = z * clamp(z + 3, 0, 6) / 6
- g = sigmoid(alpha * m + beta)
- y = x * g

Goal
- Fuse ops to reduce memory traffic and kernel launches; target ≥1.30x speedup.
