Patch for exception message expectations

[marcioAlmada]

Changes:

• adds @expectedExceptionMessageRegExp annotation
• adds TestCase::setExpectedExceptionRegex($exceptionClass, $pattern, $code) so regex expectations for exception messages are still first class API citizens
• recovers old behavior of @expectedExceptionMessage, keeps the tests introduced by Regex aware @expectedExceptionMessage #1264
• for good or bad, both @expectedExceptionMessage and @expectedExceptionMessageRegExp can now be used simultaneously
• removes ambiguity already discussed in PHPUnit 4.2: Exceptions with chars used by regular expression can break tests #1328

Relates to: #1266 #1264 #1263

Things not exactly related to this PR that I think we could do better:

• I particulary don't like the redundancy of TestCase::setExpectedException vs TestCase::setExpectedExceptionRegex, but it was the only sane way to keep the feature since there are no granular methods like TestCase::setExpectedExceptionMessage or TestCase::setExpectedExceptionCode. Any ideas?
• The way PHPUnit grabs annotation from doc blocks is spread across multiple methods at Util\Test class, and it can be kinda tricky to expand the API (look at how much edits I had to do just to achieve a minor tweak). There are many annotations libraries around, I recommend to pick one ;). I could make a PR to improve this in the future.

[whatthejeff]

Thanks again, @marcioAlmada!

[marcioAlmada]

Thank you too. Hope we can move to #1356 now.

[sun]

Thanks a lot @marcioAlmada!

Just a minor note: Would be great to squash/fixup commits into 2-3 (next time), so as to have a clean history in the mainline. If someone needs to figure out why the code works like it does in the future, they'll highly appreciate it 😉

[marcioAlmada]

@sun It tend to add a failing and a passing commit for each step I take, specially when I'm doing a PR which could potentially become a work in progress PR (when history needs to be exposed gradually), but I would have no problems if commits were squashed by someone else (@whatthejeff) since I usually delete my feature branch and pull from master later.