Text this: Rank-one approximation of a higher-order tensor by a Riemannian trust-region method