Floating-point rounding

Also written: floating-point, scientific notation

Tiny errors from storing decimals in binary, such as `-5.55e-17` where you expected 0. Compare decimals with `np.allclose`, never `==`.