Quantum state tomography: Mean squared error matters, bias does not