Corrected a bug on convolve.F

[arango]

Removed the initialization of the flag LwrtTime around line 460 of convolve.F:

        IF (Nrec(ng).gt.3) THEN
          LwrtTime(ng)=.TRUE.
          DO irec=1,Nrec(ng)-1
            ...
          END DO
        END IF

This forced ad_wrt_his.F to over-write the ocean_time in the NetCDF file with the incorrect values. All values are the same for all records since ForceTime(ng) is always the same ad the initialization time (last one processed). This also affected the impulse time management in TLF NetCDF file in weak constraint applications for 4D-PSAS and R4D-Var.

The LwrtTime switch needs to be FALSE when convolving all the record in the adjoint history file.

Many thanks to Andy Moore and Emilie for bringing this to my attention