allow estimates to be correlated

[danielwilhelm]

• currently, csranks() takes arguments the vectors x and sd, the estimates and their standard deviations, assuming that the elements of x are not correlated
• the bootstrap draws generated in csranks() are from a normal with mean zero and covariance matrix that has diagonal equal to sd and all other elements equal to 0
• we want to allow for correlation
• this can be done by passing as an argument the covariance matrix of x instead of sd
• then, the bootstrap draws generated in csranks() must be from a normal distribution with mean zero and that covariance matrix (instead of a diagonal the diagonal matrix with sd on the diagonal)

[MrDomani]

Ready on branch dev. Pull will be requested soon.