fixes for newer v8

[icambron]

The latest v8 (current Chrome and Node 25) introduces a few nasty changes:

1. more alternative whitespace characters used in formatting by v8. This time it's not just nbsp; there's also now U+202F ("narrow non-breaking space"). This affects Luxon in two ways: a) the fromFormat parsing code will now expect those new kinds of spaces, and b) the test code that makes assertions on the the output of toFormat should ignore different-kind-of-whitespace differences.

2. There's a bug in v8 (or maybe the ICU data) that affects the coptic calendar. It seems to provide "AM" and "PM" as era values. That's weird and I didn't debug it further, since it's easy to reproduce directly in intl. Instead, I switched the tests that use coptic to instead use buddhist I misunderstood the output here. I undid the changes related to this.

3. The th locale (Thai) no longer produces a localized meridiem string. I've simply removed the uses of it from the tests.

You'll notice there's only one "real" code change: item 1a. Before we handled the NBSP character by replacing any whitespace literals in the token regexes with a character class that included both a regular space and the nbsp character (i.e. [ ${nbsp}]). I've switched it to replace anything matching whitespace character in the regex with the whitespace character class, so for example, /foo bar/ will become /foo\sbar/. I think this is the correct fix, and likely what we should have done all along.

The other changes are all test-related

[diesieben07]

Hi and thanks for this!
I am currently in the process of working on V4 (which I have honestly been procrastinating on, which is a shame), which is why it took so long for me to look at this.
I think I already have something similar in my V4 branch, but this looks cleaner. I will revisit this once V4 is in a state where the code is actually useful.