Automatically disable prescribed boundary conditions

GPU accelerated models still run prescribed boundary conditions on the CPU. This is done to make it easy to use prescribed boundaries. However, this has performance penalties, since the `extboundary` data needs to be updated on the host before filling in prescribed boundaries and pushed back to the device after filling prescribed boundaries.

On meshes that do not use prescribed boundaries anywhere, or are explicitly time independent boundaries, we should enforce disabling of the prescribed boundary conditions on behalf of the user to automatically promote optimal performance. When prescribed boundary conditions are detected, then the prescribed boundary condition evaluation should be retained.