Fix a few edge cases for XRRigidTransform

First, for sufficiently large DOMPointInit Values, it was impossible to
normalize the quaternion. This now throws an Invalid State error.

Second, for sufficiently large Translation values, the DCHECK for
IsInvertible, was failing. This triggered tests/debug builds, while on
production builds, it simply returned identity. This DCHECK has been
removed and replaced with a DLOG(ERROR) and an explicit return of an
identity inverse, to match the existing behavior while spec discussions
are ongoing as far as what to do in this case. It is worth noting, that
the "matrix" attribute that the page receives in cases such as this,
contains "invalid" (e.g. Infinity) values (indeed even before reaching
this case), due to the type mismatch, where DOMPointInit/the internal
representation are doubles, and the "matrix" attribute is a
Float32Array.

Bug: 1258611
Change-Id: I572ec27a4c176a0a1a7352c8cbdf92070dcf1080
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/3257117
Commit-Queue: Alexander Cooper <[email protected]>
Commit-Queue: Klaus Weidner <[email protected]>
Auto-Submit: Alexander Cooper <[email protected]>
Reviewed-by: Klaus Weidner <[email protected]>
Cr-Commit-Position: refs/heads/main@{#937665}

webxr/xrRigidTransform_constructor.https.html

@@ -134,6 +134,11 @@
       coordDict([1.0, 2.0, 3.0, 1.0]),
       coordDict([0, 0, 0, 0])
   ), "Constructor should throw InvalidStateError for non-normalizeable orientation values");
+
+assert_throws_dom("InvalidStateError", () => new XRRigidTransform(
+    coordDict([1.0, 2.0, 3.0, 1.0]),
+    coordDict([-1.7976931348623157e+308, 0, 0, 0])
+), "Constructor should throw InvalidStateError for non-normalizeable orientation values");
   resolve();
 });