fix wrong ENTRYPOINT in Dockerfile.cuda

[guyoun]

Adjust the order and format of inputs to libvmaf_cuda according to https://ffmpeg.org/ffmpeg-filters.html#toc-libvmaf_005fcuda

The libvmaf_cuda filter allows only AV_FIX_FMT_CUDA but NV12(sw_format) is not supported.

https://github.com/FFmpeg/FFmpeg/blob/master/libavfilter/vf_libvmaf.c#L821C26-L821C41

So the current config raises [libvmaf @ 0x7f95d00050c0] Unsupported input format: nv12 like below.

Stream mapping:
  Stream #0:0 (h264) -> libvmaf_cuda
  Stream #1:0 (h264) -> libvmaf_cuda
  libvmaf_cuda:default -> Stream #0:0 (wrapped_avframe)
Press [q] to stop, [?] for help
[libvmaf @ 0x7f95d00050c0] Unsupported input format: nv12
[Parsed_libvmaf_cuda_0 @ 0x7f95d0004fc0] Failed to configure output pad on Parsed_libvmaf_cuda_0
[fc#0 @ 0x55b0b17da800] Error reinitializing filters!
[fc#0 @ 0x55b0b17da800] Task finished with error code: -22 (Invalid argument)
[fc#0 @ 0x55b0b17da800] Terminating thread with return code -22 (Invalid argument)
[vost#0:0/wrapped_avframe @ 0x55b0b181a8c0] Could not open encoder before EOF
[vost#0:0/wrapped_avframe @ 0x55b0b181a8c0] Task finished with error code: -22 (Invalid argument)
[vost#0:0/wrapped_avframe @ 0x55b0b181a8c0] Terminating thread with return code -22 (Invalid argument)
[out#0/null @ 0x55b0b181be80] Nothing was written into output file, because at least one of its streams received no packets.
frame=    0 fps=0.0 q=0.0 Lsize=       0KiB time=N/A bitrate=N/A speed=N/A

Conversion failed!

[nilfm99]

Thanks @guyoun for the contribution.

FYI @kylophone and @gedoensmax, this looks correct to me based on documentation but can't verify it now.

[gedoensmax]

For 420 video this is the correct command. 444 for example would work with the previous one already.

[nilfm99]

I think 420 video is more common, so it makes sense to have the Dockerfile be able to handle this.

[nilfm99]

LGTM, will wait a bit to merge in case there are other opinions