Custom video and audio sample resources

[Vanilagy]

This PR adds VideoSampleResource and AudioSampleResource, a way to create a video or audio sample that's backed by a custom data resource not covered by the base WebCodecs API. The intention is that you can use this to efficiently connect Mediabunny with other sample sources, for example a sample backed by an FFmpeg AVFrame, without needing to copy any data or convert to Uint8Array first.

Things to figure out:

• allocationSize and copyTo can get very complex to implement properly. Is the burden pushed fully onto the implementor of the resource, or does Mediabunny help out? I feel both must be possible: For max efficiency, the resource should be able to handle it itself, but if it's lazy, it should be able to communicate that too.
• What about drawing to a canvas? Right now, it takes the VideoFrame route, which is okay in browser environments. But, how does this extend to server environments where the canvas API needs to be polyfilled?
• Is the API good to use? Needs to be tested with an actual production-ready example.