(re-)enable after each hooks to fail tests.

In version 1.2.1, support was added to fail test in the after each hook:
> * Added `this.test.error(err)` support to after each hooks. Closes mochajs#287

While migrating from teaspoon-mocha to karma-mocha, we discovered that
in the current mocha version this isn't supported anymore (or we don't
understand how). In teaspoon this behaviour seems to be patched (running
it with the exact same mocha.js 2.2.4 file, teaspoon and karma have
different behaviour), although I couldn't quickly pinpoint the code that
does that.

This patch adds this behaviour back to the code base, but also seems to
break certain behaviour (3 tests are failing with this pull request).
But I don't understand what those test failures exactly mean. Also I
found it quite difficult to add test coverage of this behaviour. Since
there aren't any tests calling `runTests`.

Help in figuring out what we've done wrong or how we could fix this pull
request, would be appreciated. Not supporting this behaviour anymore due
to added complexity in the code, would also be fine, but adding some
documentation about dropping this functionality would be nice.

lib/runner.js

@@ -544,9 +544,16 @@ Runner.prototype.runTests = function(suite, fn) {
         }
 
         test.state = 'passed';
-        self.emit('pass', test);
-        self.emit('test end', test);
-        self.hookUp('afterEach', next);
+        // Run the actual emit that the test passed after the forEach hooks,
+        // so they can still fail a test. And it's not already reported as
+        // passed.
+        self.hookUp('afterEach', function() {
+          if (test.state === 'passed') {
+            self.emit('pass', test);
+          }
+          self.emit('test end', test);
+          next();
+        });
       });
     });
   }