Several bugs in diaptr

[hadcv]

1. timing_stop is not called if l_diaptr == .FALSE.
2. The "zosrf" diagnostic outputs the wrong data (the 3D surface area, instead of the zonal integral of this quantity) because the result is only defined on one i row.
3. btmsk is undefined if "subbasins.nc" does not exist

Fix

1. Call timing_stop in the l_diaptr == .FALSE. case:

​http://forge.ipsl.jussieu.fr/nemo/browser/NEMO/trunk/src/OCE/DIA/diaptr.F90?rev=13497#L101

- IF( .NOT. l_diaptr )   RETURN
+ IF( .NOT. l_diaptr ) THEN
+    IF( ln_timing ) CALL timing_stop('dia_ptr')
+    RETURN
+ ENDIF

2. Set the zonal integral over the full working array as in other cases:

​http://forge.ipsl.jussieu.fr/nemo/browser/NEMO/trunk/src/OCE/DIA/diaptr.F90?rev=13497#L199

  DO jn = 1, nptr
     zmask(1,:,:) = ptr_sjk( zmask(:,:,:), btmsk(:,:,jn) )
+    DO ji = 1, jpi
+       zmask(ji,:,:) = zmask(1,:,:)
+    ENDDO
     z4d1(:,:,:,jn) = zmask(:,:,:)
  ENDDO
  CALL iom_put( 'zosrf', z4d1 )

3. Initialise btmsk:

​http://forge.ipsl.jussieu.fr/nemo/browser/NEMO/trunk/src/OCE/DIA/diaptr.F90?rev=13497#L355

+ btmsk(:,:,:) = 0._wp
  btmsk(:,:,1) = tmask_i(:,:)

[smasson]

In 13557:

Error: Failed to load processor CommitTicketReference

No macro or processor named 'CommitTicketReference' found

[smasson]

I reported points 1 and 2.
I agree with the 3rd point but with Christian we decided to correct it in a different way as we now allow the number of basin to be 1 or 5