When experimenting with test/studies/shootout/thread-ring, it was discovered
that using a coforall loop + begin statements to start tasks was faster than
using only a coforall loop, the latter of which performed roughly similar to
just using a for loop with begin statements.  This case is not easily
demonstrated using trivial code (i.e. the problem becomes noticable only when
the tasks being performed are more than just spinning for an arbitrary amount
of time), so the thread-ring code has been copied for the purpose of this test.
(accurate as of revision 21783)
