S2LatLngRect::GetCentroid test failures

[jmr]

A few years ago, the comparison threshold for rect centroid splitting was increased from 1e-15 to 2e-15.

s2geometry/src/s2/s2latlng_rect_test.cc

Lines 646 to 649 in ac448e2

	// About 3.5% flaky with 1e-15, so increase to 2e-15, which is <0.1% flaky.
	EXPECT_LE(
	(r.GetCentroid() - child0.GetCentroid() - child1.GetCentroid()).Norm(),
	2e-15);

Now I notice that the "full latitude range" tests sometimes fail with 1e-15 and need ~4/3 e-15.

s2geometry/src/s2/s2latlng_rect_test.cc

Line 684 in ac448e2

EXPECT_NEAR(0.25 * M_PI * sin(alpha) / alpha * r.Area(),

@ericveach Do you remember if these were supposed to be rigorous, or just a small value that seemed to work?