Lenient stackTraceLimit works across browsers

[erights]

Fixes #525

Using #590 to iterate within several browsers, I repaired several issues.

• Fix Error.stackTraceLimit assignment on Firefox #525 was that assigning Error.stackTraceLimit was failing in non-v8-based browsers because Error had no such property and is frozen. Although the property is indeed non-standard and absent in non-v8 systems, Google's documentation encourages assignments to it anyway. This is a noop elsewhere, but is harmless when Error is unfrozen. To accommodate this practice, this PR installs a noop Error.stackTraceLimit accessor property on every platform. It is a noop everywhere except in the start compartment of a v8-based platform.
• On the Safari browser (and presumably on any host using the JSC engine) DataView has a non-standard but non-deletable BYTES_PER_ELEMENT property whose value is a number. Since it is non-deletable but harmless, we added it to our whitelist.
• We added logic for obtaining the stack string for FF (using the proposed Error.prototype.stack accessor property) and others including Safari using the error.stack` property if present. Otherwise (such as on XS currently) the default stack remains the empty string.

In the process, made the console output more pleasant for the common case of console messages containing only one error argument. For this case, nothing is gained by nesting the error within a group. Removing this extra visual noise helps this common case.

Because I built it using #590 it was easy to build it on #590 , which is fine if #590 gets merged first. If #590 gets delayed, I can make this independently mergeable since it has no logical dependence on #590.

[dckc]

At a glance, I don't see anything wrong, but I'm not confident I could maintain this, so I'm inclined to leave it to others.

[warner]

Looks pretty good to me, but I'll defer to @michaelfig or @kriskowal as the real experts.

[michaelfig]

LGTM.