zero out I_q when castro.drive_initial_convection resets state

[zingale]

When we do SDC with castro.drive_initial_convection=1, every time we hit castro.drive_initial_convection_reinit_period we reset the fluid state, but keep the velocities. This means that in the next timestep, the energy generation will change discontinuously, so using the lagged I_q from the previous step across this reset is not a good idea. We should reset I_q in this case.