TexSubImage2D obtains the full image pixels then uses a subset

[jdm]

The get_image_pixels helper obtains the entire image pixels. I could imagine this being particularly wasteful when we fix #21995, where the video frames are significantly larger that the content that is displayed at any given moment. It would be wise to support obtaining only the necessary pixels for a given rectangle instead.

[nox]

#22225 makes that code use a IpcSharedMemory, which means that for things such as HTMLImageElement we don't need to copy any pixel at all in this method. The same should be done for HTMLVideoElement.

I'm not sure what you mean that video frames are larger than the content they are displayed in.