EllipseGeometryLibrary's computeEllipsePositions misses critical points

[nmschulte-aviture]

Because the logic for computing the positions about the ellipse works by computing the "north" half of the ellipse (by computing the "west" half and then its reflection) and then its reflection, with low granularities, say a granularity of PI_OVER_TWO (to retrieve the four critical points, 0, pi/2, pi, 3pi/4), all of the resolution is focused around the west/east points, and none about the north/south points. (In fact, with PI_OVER_TWO, the two east/west points are duplicated in the calculation).

You can see this visually in the screenshot from @bagnell's work here re: granularity:
#2283 (comment)

I'm wondering if simply adding the calculation for the critical points would be enough to resolve this? Perhaps changing the math to mirror about the 45, 135, 225 and 315 degree angles (rather than the current 0/90/180/270) would help in conjunction w/ adding the critical points. This seems like it would always ensure an accurate representation of the ellipse from the computation.