bug in ice-ocean stress

Some (long) time ago, a bug has been introduced by mistake in the ice-ocean stress when using the implicit drags. It impacts both v4.2 and the main and may have dramatic impact on ice-ocean interactions with large consequences for the AMOC (to be verified).

Solution: replace these lines:

zutau_ice   = 0.5_wp * tmod_io(ji,jj)   &
     &                 * ( ( u_ice(ji,jj) + u_ice(ji-1,jj) ) - ( pu_oce(ji,jj) + pu_oce(ji-1,jj) ) )   ! add () for
zvtau_ice   = 0.5_wp * tmod_io(ji,jj)   &
     &                 * ( ( v_ice(ji,jj) + v_ice(ji,jj-1) ) - ( pv_oce(ji,jj) + pv_oce(ji,jj-1) ) )   ! NP repro

by

zutau_ice   = 0.5_wp * tmod_io(ji,jj)   &
     &                 * ( ( u_ice(ji,jj) + u_ice(ji-1,jj) ) - zflagi * ( pu_oce(ji,jj) + pu_oce(ji-1,jj) ) )   ! add () for
zvtau_ice   = 0.5_wp * tmod_io(ji,jj)   &
     &                 * ( ( v_ice(ji,jj) + v_ice(ji,jj-1) ) - zflagi * ( pv_oce(ji,jj) + pv_oce(ji,jj-1) ) )   ! NP repro