VERY IMPORTANT: Corrected bug in MPDATA

[arango]

In ​src:ticket:839 coupled of bugs were introduced to the MPDATA tracer advection in routine step3d_t.F. The following I-loop the wrong range:

!
!  If MPDATA, time-step horizontal advection for intermediate diffusive
!  tracer, Ta. Otherwise, time-step horizontal advection term.
!  Advective fluxes have units of Tunits m3/s.  The new tracer has
!  units of m Tunits.
!
          IF (Hadvection(itrc,ng)%MPDATA) THEN
            DO j=JstrVm2,Jendp2i
              DO i=IstrUm2,IstrUm2
                ...
                ...
              END DO
            END DO

We need to have instead:

           IF (Hadvection(itrc,ng)%MPDATA) THEN
             DO j=JstrVm2,Jendp2i
               DO i=IstrUm2,Iendp2i
                 ...
               END DO
             END DO

The I-range needs to be IstrUm2:Iendp2i. It is a copying and pasting typo. Many thanks to Chuning Wang for ​reporting this bug.

The second bug is due to a combination of CPP option and logical switch in the vertical diffusion term when SPLINES_VDIFF is activated. If MPDATA, the conditional logic is wrong, and the vertical diffusion term is not included. We need to have instead:

!
!-----------------------------------------------------------------------
!  Time-step vertical diffusion term.
!-----------------------------------------------------------------------
!
      J_LOOP2 : DO j=Jstr,Jend                  ! start pipelined J-loop
        DO itrc=1,NT(ng)
          ltrc=MIN(NAT,itrc)

# ifdef SPLINES_VDIFF
          IF (.not.((Hadvection(itrc,ng)%MPDATA).and.                   &
     &              (Vadvection(itrc,ng)%MPDATA))) THEN
            ...
          ELSE
# endif
            ...

# ifdef SPLINES_VDIFF
          END IF
# endif
        END DO
      END DO J_LOOP2

Again, many thanks to Chuning Wang for ​reporting for bringing this logic problem to my attention. There is a wrong combination of the CPP option and IF-conditional.

Made similar changes to the TLM, RPM, ADM versions of step3d_t.F. Luckily, the MPDATA is not supported in the adjoint-based algorithms.