WebXRHand doesn't efficiently obtain hand poses

[rgerd]

The WebXR hand tracking API explainer recommends using fillPoses and fillJointRadii as a way of reducing the number of javascript objects created per-frame while updating tracked hand models:
https://github.com/immersive-web/webxr-hand-input/blob/main/explainer.md#efficiently-obtaining-hand-poses

WebXRHand still uses the inefficient method:

Babylon.js/src/XR/features/WebXRHandTracking.ts

Line 347 in 3a9e132

let pose = xrFrame.getJointPose!(xrJoint, referenceSpace);

In order for BabylonNative to maintain support, we will want to get BabylonJS/BabylonNative#773 closed first.

[RaananW]

Hi @rgerd - will you be implementing this? Should I assign you?

[rgerd]

@RaananW Yes I'm working on it now. Thanks!