Refactor cosine lr scheduler

[wyg1997]

• rename CosineAnnealingLR -> CosinDecayLR，对齐 lazy 的 CosineDecay 接口
• 实现对齐 torch 的 CosineAnnealingLR 接口

[github-actions]

CI failed, removing label automerge

[github-actions]

Speed stats:

GPU Name: GeForce GTX 1080 

PyTorch resnet50 time: 139.7ms (= 6986.4ms / 50, input_shape=[16, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 128.1ms (= 6404.6ms / 50, input_shape=[16, 3, 224, 224], backward is enabled)
Relative speed: 1.09 (= 139.7ms / 128.1ms)

PyTorch resnet50 time: 82.0ms (= 4098.2ms / 50, input_shape=[8, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 74.5ms (= 3723.4ms / 50, input_shape=[8, 3, 224, 224], backward is enabled)
Relative speed: 1.10 (= 82.0ms / 74.5ms)

PyTorch resnet50 time: 57.6ms (= 2878.0ms / 50, input_shape=[4, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 48.2ms (= 2407.8ms / 50, input_shape=[4, 3, 224, 224], backward is enabled)
Relative speed: 1.20 (= 57.6ms / 48.2ms)

PyTorch resnet50 time: 48.8ms (= 2438.3ms / 50, input_shape=[2, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 41.8ms (= 2090.9ms / 50, input_shape=[2, 3, 224, 224], backward is enabled)
Relative speed: 1.17 (= 48.8ms / 41.8ms)

PyTorch resnet50 time: 40.0ms (= 1998.0ms / 50, input_shape=[1, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 39.7ms (= 1984.5ms / 50, input_shape=[1, 3, 224, 224], backward is enabled)
Relative speed: 1.01 (= 40.0ms / 39.7ms)