When a ~= 0, the 2nd derivative of x/sqrt(x^2+a^2) is positive for negative x, and negative for positive x. Hence it is neither convex nor concave. Therefore, it can’t be used in CVX.
Of course, we don;t know what you are trying to do. For instance if you are trying to model a constraint x/norm([x a]) >= 1, you can multiply that out to x >= norm([x a]), which is a convex constraint, and is accepted by CVX.
I commend you to carefully read