Preserve sign of -0 in arbitrary_precision mode (#1346, thanks @yuxi-liu-wired) Fix comparison of large non-integer values with f32::INFINITY (#1347, thanks @yuxi-liu-wired)