Nan Error Handling

When a matrix has one or more negative eigenvalues, this leads to a complex number and nan result later in the calculation. The most common source of the error is a lack of sampling in the trajectory. We need error handling to notice when this happens and give a clear message to the user so that they know what went wrong and have a clue as to what needs to be done to fix the problem.