Fill correct stack info in pthread_getattr_np stub

[hoodmane]

See use case here:
https://github.com/python/cpython/pull/130398/files#diff-c22186367cbe20233e843261998dc027ae5f1f8c0d2e778abfa454ae74cc59deR365-R380

I guess for stack protection, this only enables introspecting whether we run out of the spill stack. I'm not sure whether to expect running out of spill stack or running out of true stack to be more common. But at least this unbreaks the code.

[hoodmane]

The closure tests are now failing for me locally with:

building:ERROR: Closure compiler run failed:

building:ERROR: /tmp/emtest_aihv0i03/emscripten_temp_z4b7qfms/a.out.js:121:4: ERROR - [JSC_UNDEFINED_VARIABLE] variable assert is undeclared
  121|     assert(typeof data == 'object');

[sbc100]

I experimented with using static struct initializer instead but it looks like it actually caused and overall regression: #23904

[sbc100]

Nice! LGTM

[hoodmane]

I guess this is still a few bytes bigger than before because we added pthread_getattr_np.c.

[sbc100]

Seems like a 40 byte increase is more than I would expect from this change.. I was hoping for more like 6 bytes (2 bytes each for the 3 extra store instructions).

[hoodmane]

Well we also added pthread_getattr_np.c which looks like:

int pthread_getattr_np(pthread_t t, pthread_attr_t *a)
{
	*a = (pthread_attr_t){0};
	a->_a_detach = t->detach_state>=DT_DETACHED;
	a->_a_guardsize = t->guard_size;
	a->_a_stackaddr = (uintptr_t)t->stack;
	a->_a_stacksize = t->stack_size;
	return 0;
}

Presumably the other 34 bytes come from that?

[sbc100]

But that file should not linking into any of those tests. It should only be linked into the (rare) programs that use it.