fix upsample bug

[BBuf]

修复upsample.h中的一个BUG。

[github-actions]

CI failed, removing label automerge

[github-actions]

Speed stats:

GPU Name: GeForce GTX 1080 

PyTorch resnet50 time: 139.7ms (= 6986.3ms / 50, input_shape=[16, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 127.9ms (= 6392.8ms / 50, input_shape=[16, 3, 224, 224], backward is enabled)
Relative speed: 1.09 (= 139.7ms / 127.9ms)

PyTorch resnet50 time: 84.2ms (= 4210.3ms / 50, input_shape=[8, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 74.5ms (= 3725.7ms / 50, input_shape=[8, 3, 224, 224], backward is enabled)
Relative speed: 1.13 (= 84.2ms / 74.5ms)

PyTorch resnet50 time: 60.3ms (= 3016.1ms / 50, input_shape=[4, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 49.0ms (= 2451.9ms / 50, input_shape=[4, 3, 224, 224], backward is enabled)
Relative speed: 1.23 (= 60.3ms / 49.0ms)

PyTorch resnet50 time: 47.5ms (= 2377.5ms / 50, input_shape=[2, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 41.0ms (= 2047.8ms / 50, input_shape=[2, 3, 224, 224], backward is enabled)
Relative speed: 1.16 (= 47.5ms / 41.0ms)

PyTorch resnet50 time: 43.3ms (= 2163.9ms / 50, input_shape=[1, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 43.7ms (= 2187.4ms / 50, input_shape=[1, 3, 224, 224], backward is enabled)
Relative speed: 0.99 (= 43.3ms / 43.7ms)

[github-actions]

Speed stats:

GPU Name: GeForce GTX 1080 

PyTorch resnet50 time: 140.8ms (= 7039.3ms / 50, input_shape=[16, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 128.0ms (= 6399.1ms / 50, input_shape=[16, 3, 224, 224], backward is enabled)
Relative speed: 1.10 (= 140.8ms / 128.0ms)

PyTorch resnet50 time: 84.2ms (= 4208.3ms / 50, input_shape=[8, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 74.5ms (= 3723.1ms / 50, input_shape=[8, 3, 224, 224], backward is enabled)
Relative speed: 1.13 (= 84.2ms / 74.5ms)

PyTorch resnet50 time: 60.3ms (= 3017.2ms / 50, input_shape=[4, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 49.0ms (= 2452.2ms / 50, input_shape=[4, 3, 224, 224], backward is enabled)
Relative speed: 1.23 (= 60.3ms / 49.0ms)

PyTorch resnet50 time: 46.6ms (= 2332.3ms / 50, input_shape=[2, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 41.4ms (= 2072.1ms / 50, input_shape=[2, 3, 224, 224], backward is enabled)
Relative speed: 1.13 (= 46.6ms / 41.4ms)

PyTorch resnet50 time: 40.6ms (= 2031.9ms / 50, input_shape=[1, 3, 224, 224], backward is enabled)
OneFlow resnet50 time: 41.3ms (= 2063.3ms / 50, input_shape=[1, 3, 224, 224], backward is enabled)
Relative speed: 0.98 (= 40.6ms / 41.3ms)

[PragmaTwice]

This fix has accually no effect.
A will be passed to the second parameter typed const T of cubic_convolution1<T> and cubic_convolution2<T>, and obviously an implicit cast will be performed due to T = int.