use PointerEvents instead of mouse events so we can use setPointerCapture

[leon]

Component your feature request relates to

• Signalling server
• Frontend library
• Frontend UI library
• Matchmaker
• SFU
• This repo

Problem your feature solves? Please describe.

#848
#349

There was a recent PR merge which was supposed to solve the problem of tracking the mouse while outside the window / viewport.
but it did not solve it, because to track the mouse outside the browser window you need to use the native pointer events setPointerCapture passing in the pointerId.

Describe the solution you'd like
I did a test implementation where we go over to use pointer events in the MouseControllerHovering

master...leon:PixelStreamingInfrastructure:pointerevents-with-capture

I think this could solve the problem.
pointer events is a drop in replacement for mouse and touch, but since we are handling the touch events separatly, I made this one to only work with mouse and pen.

So give it a go an see what you think :)

[mcottontensor]

Thanks for digging into this.

If you want to turn the branch into a PR, there are a few things I'd like to sort out first:

1. There's a conflict between the { passive: true } listeners and preventDefault(). pointermove and wheel are registered passive, but onPointerMove and onMouseWheel both still call preventDefault(), which passive listeners can't do. That will log warnings and, for wheel, let the page scroll while the cursor is over the stream. Either drop passive or drop the preventDefault on those.

2. Multi-button / chorded presses need handling. For mouse, pointer events only fire pointerdown on the first button and pointerup on the last release, with intermediate button changes arriving as pointermove and a changed buttons bitmask. Since onPointerDown/onPointerUp key off event.button, a second button pressed while the first is held won't send a MouseDown, and releasing one of two held buttons won't send a MouseUp. UE expects paired per-button down/up, so we'd want to diff event.buttons in the move handler to synthesize the intermediate ones. Worth testing with something that uses two buttons at once.

3. For cleanup, pointercancel is covered, but it'd be good to also handle lostpointercapture and make sure the MouseUp we synthesize on a cancel can't leave the engine thinking a button is still down.

4. Minor: could you keep the fuller double-click comment (including the #10 reference)? It explains why the auto-release exists, and I'd rather not lose that context.

Get those tidied up and we'd be happy to review it. Thanks again for the solid investigation.