Fix acqusistion function bug when std output by GP is 0.0

[csbo98]

In order to calculate the acquisition function EI and PI for a given point x, we will use mean and std output by the gaussian process. However, std may be 0.0(e.g., the gaussian process will set std to 0.0 when we sample a point x that has been sampled before), which will cause an error when calculating EI and PI. In function _ei, when std is 0.0, we can't calculate the EI value using a * norm.cdf(z) + std * norm.pdf(z), instead, we should set the EI value to 0.0. So does function _poi. Therefore, I changed function _ei and _poi to fix this bug.
I refer to Acquisition functions in Bayesian Optimization, which discussed the acquisition function in detail and explained the derivation process of PI and EI

[bwheelz36]

Do you have an example script where this error is occurring?