Let us assume that that the Lagrangian is polynomial in the velocities $\dot{q}_i$. A quadratic term in the velocities $\dot{q}_i$ would properly be thought of as part of the kinetic energy rather than the potential energy. And terms that are higher order in the velocities are equivalent (via integration by parts) to terms that involve higher-order derivatives of $q_i$, which will generically lead to Ostrogradski instabilities, so it is also reasonable to eliminate those from consideration.
Let us also assume that the Lagrangian is independent of $t$. This ensures the existence of a conserved quantity that can be identified as the energy of the system.
But this means that the only option left is for $U$ to be quasi-linear in $\dot{q}_i$, i.e., of the form
$$
U(q_i, \dot{q}_i) = \sum_k A_k(q_i) \dot{q}_k + \tilde{U}(q_i)
$$
where the functions $A_k$ and $\tilde{U}$ only depend on the coordinates. This is exactly the same form as the Lorentz force (if the $q_i$ coordinates are Cartesian coordinates, this is just $\vec{A} \cdot \vec{v}$.) And the resulting force is of the form
$$
F_k = \sum_j \left[\left( \frac{\partial A_k}{\partial q_j} - \frac{\partial A_j}{\partial q_k} \right) \dot{q}_j\right] + \frac{\partial \tilde{U}}{\partial q_k}
$$
which again looks an awful lot like the Lorentz force.
So under the above assumptions, it does not seem like there are a lot of options other than "Lorentz-like" forces to arise from velocity-dependent potentials.
