feature: play/pause and running promise modified

[JoaquinBCh]

This PR adds the ability to pause and resume playback in the player.

• Modified the play() method, which toggles between pausing and resuming playback:
• On pause, it unsubscribes from the current audio and video tracks.
• On resume, it resubscribes to the audio and video tracks.
• Adjusted the handling of this.#running and the #runTrack tasks to allow resubscription after pausing.
• Can now click either the Play/Pause button or directly on the player (canvas) to toggle playback.
  

[englishm]

Thanks, @JoaquinBCh!

As mentioned on Discord, as much fun as it is to data mosh big buck bunny by pausing and playing and applying P frames to old state, I think we probably want something to ensure we restart playback on a keyframe.

bbb_datamosh.mp4

#3 added some code along these lines:

moq-js/lib/playback/worker/video.ts

Lines 122 to 141 in f6c7424

	//At the start of decode , VideoDecoder seems to expect a key frame after configure() or flush()
	if (this.#decoder.state == "configured") {
	if (this.#waitingForKeyframe && !frame.sample.is_sync) {
	console.warn("Skipping non-keyframe until a keyframe is found.")
	return
	}
	// On arrival of a keyframe, allow decoding and stop waiting for a keyframe.
	if (frame.sample.is_sync) {
	this.#waitingForKeyframe = false
	}
	const chunk = new EncodedVideoChunk({
	type: frame.sample.is_sync ? "key" : "delta",
	data: frame.sample.data,
	timestamp: frame.sample.dts / frame.track.timescale,
	})
	this.#decoder.decode(chunk)
	}

But, I think because of how this shuffles around the tasks to actually run the tracks, that code isn't working in the same way anymore here.

Once we can sort out how to ensure a cleaner restart of playback here, I'm happy to merge this in.
Thanks again for the contribution!

[nicolaslevy]

Nice! Just in case, we have to test with ./dev/pub_multi_track to see "better" the issue with the Key frames

[JoaquinBCh]

I set to true this.#waitingForKeyframe when pause

bbb_pause_play.webm.mp4

[englishm]

Tested again, and this appears to be properly ensuring we always wait and reinitialize with keyframes now. Thanks!