Clean up inconsistent hit-test behavior

[hi-liang]

1. Working raycasting transient touches in XRbrowser
2. Determine guarantees of xr select intersections in a-cursor successfully disabling ar-hit-test before it can execute the hit test, as both events firing have been regularly observed. This may involve a monkeypatch of ar-hit-test component so that it doesn't listen on select, but rather a custom event invoked from a-cursor after no intersections are detected.

It may be worth considering if this is the proper approach for reconciling real-world hit-testing vs virtual object raycast intersections: if a virtual intersection occurs, regardless of distance from the view origin, it will take priority over a hit-test plane.