Major bug fix for tendencies from NSAS cumulus scheme (#71)

TYPE: bug fix

KEYWORDS: NSAS cumulus scheme, tendencies

SOURCE: Sebastien Masson, LOCEAN - IPSL, Univ. Pierre et Marie Curie, France

DESCRIPTION OF CHANGES:
In module_cu_nsas.F, all cumulus tendency terms (*cuten variables) are wrong as they are divided by dt*stepcu (where stepcu is the number of time steps in CUDT) instead of dt while the deep (nsas2d) and shallow (nscv2d) routines are called with dt. As tendency terms are kept constant during stepcu time steps, they should not be divided by dt*stepcu but by dt.

LIST OF MODIFIED FILES:
phys/module_cu_nsas.F

TESTS CONDUCTED: reg tested. Ran January case, new changes provide more heating and less rainfall. The user verified this by playing with time_step and cudt. Any reasonable value of time_step and cudt gives very similar results when stepcu = 1, but results drastically change when stepcu = 2 or higher values.

phys/module_cu_nsas.F

@@ -160,7 +160,7 @@ subroutine cu_nsas(dt,dx,p3di,p3d,pi3d,qc3d,qi3d,rho3d,itimestep,stepcu,    &
        cu_act_flag(i,j)=.TRUE.
      enddo
    enddo
-   delt=dt*stepcu
+   delt=dt
    rdelt=1./delt
 !
 ! outer most J_loop