Specify and test how texture upload should work with video elements

[kimmok]

Specify and test how texture upload should work with video elements.

https://www.khronos.org/registry/webgl/specs/latest/1.0/#6.9

The obvious expectation would be that the texture would be the size of HTMLVideoElement.videoWidth/videoHeight.

However, texture-corner-case-videos.html does not catch this expectation.

In fact, the test was modified to pass the inverse of this, e.g. behavior where the texture might be smaller.

Also, texture-corner-case-videos seems to contain hidden errors for the texSubImage2D case. The logic is that the texture is created with HTMLVideoElement.videoWidth/videoHeight but then uploaded with texSubImage2D. The upload might be smaller than videoWidth, videoHeight, but this is not caught by the test.

So the ideal intention in #2464 was to let the browser upload video data unscaled 1:1. However, this is problematic as:

1. it is not specified how it works
2. tests about it have bugs
3. it is not possible to use it after it is specified, as the needed property (video data size) is not available in HTMLVideoElement. E.g. you can upload unscaled, but using texSubImage2D to update is impossible as we have no way of knowing what parameters pass as x,y,w,h.

Here's a small tester to print rudimentary cases of how the current browsers work:
webgl-video-element-texture-size-tester.zip

WebCodecs expose the needed data, but that behavior is not specified in WebGL either.

[kdashg]

Here's a jsfiddle with a src-less Image upload:
https://jsfiddle.net/w25r0b1k/

Behavior is different between Firefox and Chrome.

[kdashg]

I agree that it feels like we should map:

texImage2D(GLenum target, GLint level, GLint internalformat, GLenum format, GLenum type,
                       HTMLVideoElement video); {
    texImage2D(target, level, internalformat, video.videoWidth, video.videoHeight, 0, format, type, video); // webgl2 entrypoint
}