If the thread releases the join lock before the cancel lock, the thread may
sometimes still be alive at cancel_dump_tracebacks_later() exit. So the cancel
lock may be destroyed while the thread is still alive, whereas the thread will
try to release the cancel lock, which just crash.

Another minor fix: the thread doesn't release the cancel lock if it didn't
acquire it.

The split probably won't happen for a while, but I might as well lay the
groundwork now since I'll be adding new test modules before too long.

initfsencoding() was not called yet: detect bootstrap (startup) issues earlier.

called yet: detect bootstrap (startup) issues earlier.

Stop if the stack overflow doesn't occur after allocating 100 MB on the stack.

 * dump_traceback_later() => dump_tracebacks_later()
 * cancel_dump_traceback_later() => cancel_dump_tracebacks_later()

Windows.

Windows.

ti can raise an exception even if PyLong_Check() has
succeeded.

(This version is a cleaned-up backport of a fix by Senthil Kumaran.)