Corrected bug in def_dai.F

[arango]

The definition of the free-surface had an additional variable argument to include water_points attribute in the data assimilation initial/restart file (DAINAME):

!
!  Define free-surface.
!
        Vinfo( 1)=Vname(1,idFsur)
        Vinfo( 2)=Vname(2,idFsur)
        Vinfo( 3)=Vname(3,idFsur)
        Vinfo(14)=Vname(4,idFsur)
        Vinfo(16)=Vname(1,idtime)
        Vinfo(20)='mask_rho'
        Vinfo(22)='coordinates'
        Aval(5)=REAL(Iinfo(1,idFsur,ng),r8)
        status=def_var(ng, iNLM, DAI(ng)%ncid, DAI(ng)%Vid(idFsur),     &
# ifdef WET_DRY
     &                 NF_FRST, nvd3, t2dgrd, Aval, Vinfo, ncname,      &
     &                 SetFillVal = .FALSE.)
# else
     &                 NF_FRST, nvd3, t2dgrd, Aval, Vinfo, ncname)

# endif
        IF (exit_flag.ne.NoError) RETURN

The Vinfo(20)='mask_rho' attibute is wrong for this file.

Also, all the I/O NetCDF ID (ncid) is initialized to closed state in mod_iounits.F:

!
!-----------------------------------------------------------------------
!  Initialize I/O NetCDF files ID to close state.
!-----------------------------------------------------------------------
!
      DO ng=1,Ngrids
        ADM(ng)%ncid=-1
        ADS(ng)%ncid=-1
        AVG(ng)%ncid=-1
        BLK(ng)%ncid=-1
        BRY(ng)%ncid=-1
        CLM(ng)%ncid=-1
        DAI(ng)%ncid=-1
        DAV(ng)%ncid=-1
        DIA(ng)%ncid=-1
        ERR(ng)%ncid=-1
        FLT(ng)%ncid=-1
        FWD(ng)%ncid=-1
        GRD(ng)%ncid=-1
        GST(ng)%ncid=-1
        HIS(ng)%ncid=-1
        HSS(ng)%ncid=-1
        IAD(ng)%ncid=-1
        INI(ng)%ncid=-1
        IRP(ng)%ncid=-1
        ITL(ng)%ncid=-1
        LCZ(ng)%ncid=-1
        LZE(ng)%ncid=-1
        NUD(ng)%ncid=-1
        OBS(ng)%ncid=-1
        QCK(ng)%ncid=-1
        RST(ng)%ncid=-1
        SSF(ng)%ncid=-1
        STA(ng)%ncid=-1
        TLF(ng)%ncid=-1
        TLM(ng)%ncid=-1

        TIDE(ng)%ncid=-1

        NRM(1:4,ng)%ncid=-1
        STD(1:4,ng)%ncid=-1
      END DO