WIP: Use THREE.DeviceOrientationControls when webvrpolyfill is not used #4355

[MK-LucidWeb]

Description:
Device orientation is not taken into account anymore on Chrome m79 (related to issue #4355).

The component look-controls uses PolyfillControls

See:
look-controls.js:34
device.js:189

• In device.js, first it retrieves the VR display using the polyfill

var vrDisplay = window.webvrpolyfill && window.webvrpolyfill.getPolyfillDisplays()[0];

• Then in PolyfillControls#update, a check to see if there is a vr display. If there is none, it short circuits.

if (!vrDisplay) { return; }

• As the polyfill is not used in Chrome m79, it always short circuits.

This PR tries to change the working logic as little as possible.

Changes proposed:

• Import super-three/examples/js/controls/DeviceOrientationControls into THREE namespace
• look-controls: Check if we are using webpolyfillvr
• look-controls: if there is no webpolyfillvr, then fallback on THREE.DeviceOrientationControls

Questions:

• Should we change the name of polyfillControls and polyfillObject?
• Is checking for window.webvrpolyfill presence enough?

[MK-LucidWeb]

This needs thorough testing, imho.

Edit:// also checking why the tests are failing
Edit2:// seems like it messes with the initial camera rotation. This needs to be fixed.

[dmarcos]

Thanks, it makes sense to use DeviceOrientationControls for magic window mode. I took your logic, made some minor changes and used DeviceOrientationControls without the DeviceOrientationEvent.requestPermission logic that we handle separately. Congrats on your first contribution 🥇

[dmarcos]

Close in favor of #4363

@mako-lw I tested on Android and iOS devices and it's working my side. Let me know if the issue is resolved. We can adjust if necessary.