Resize relies on interpolate's native uint8 handling

[vfdev-5]

Description:

• Now that Add uint8 support for interpolate for CPU images pytorch#90771 is merged, let Resize() rely on interpolate()'s native uint8 handling instead of converting to and from float.

  • uint8 input is not casted to f32 for nearest and bilinear modes

Context: #7217

Benchmarks:

[-- Resize cpu torch.uint8 InterpolationMode.NEAREST -]
                         |  resize v2  |  resize stable
1 threads: --------------------------------------------
      (3, 400, 400)      |      462    |        492
      (16, 3, 400, 400)  |     6870    |      10100

Times are in microseconds (us).

[- Resize cpu torch.uint8 InterpolationMode.BILINEAR -]
                         |  resize v2  |  resize stable
1 threads: --------------------------------------------
      (3, 400, 400)      |      330    |        853
      (16, 3, 400, 400)  |     4480    |      14900

Times are in microseconds (us).

[-- Resize cpu torch.uint8 InterpolationMode.BICUBIC -]
                         |  resize v2  |  resize stable
1 threads: --------------------------------------------
      (3, 400, 400)      |      1.7    |        1.7
      (16, 3, 400, 400)  |     28.6    |       28.7

Times are in milliseconds (ms).

Source

[vfdev-5]

Closing in favor of #7557