pub fn expm1<F: Float>(x: F) -> F
e^x - 1, without the loss of digits near 0 that exp(x) - 1 has.
exp(x) - 1