split of dynamic-top-change-005

Here is a diff for two *additional* tests that split
dynamic-top-change-005 into two tests, with the intent that we add
these additional tests without removing the original, so that we can
make a better argument about the failures being independent bugs,
and therefore argue that the lack of two independent implementations
is not a sign of lack of implementability.

(In particular, WebKit should still fail both halves of the split,
but Opera should pass 005a (I tested this), IE9 should pass 005b (I
did not test this), and Gecko should pass both in addition to the
original.)

Since this test is in the approved/ folder, I believe I need
reviews.

-David

-- 
L. David Baron                                 http://dbaron.org/
Mozilla Corporation                       http://www.mozilla.com/