Fix rescaling of internal tide of debugging code

  Corrected the internal tide rescaling arguments so that some of the debugging
variables have units that are consistent with their documented units.  This
involved changing the scale arguments to global_area_integral to tmp_scale
arguments in 8 places so that the units of the output retain the scaling of the
input variable.  The multiplication by the reverse of the scaling factor was
also added to 4 debugging output messages.

  The documented units of internal wave energies in 5 register_restart_field
calls were previously given as "m3 s-2" in non-Boussinesq mode and "J m-2" in
Boussinesq mode when the reverse was actually true.  This has now been
corrected.

  Also replaced the scale argument to 35 chksum calls with the equivalent but
preferred unscale argument, following the pattern elsewhere in the MOM6 code.

  All answers are bitwise identical, and only debugging code was modified.

src/parameterizations/lateral/MOM_MEKE.F90

@@ -272,7 +272,7 @@ subroutine step_forward_MEKE(MEKE, h, SN_u, SN_v, visc, dt, G, GV, US, CS, hu, h
       if (allocated(MEKE%mom_src)) &
         call hchksum(MEKE%mom_src, 'MEKE mom_src', G%HI, unscale=US%RZ3_T3_to_W_m2*US%L_to_Z**2)
       if (allocated(MEKE%mom_src_bh)) &
-        call hchksum(MEKE%mom_src_bh, 'MEKE mom_src_bh', G%HI, scale=US%RZ3_T3_to_W_m2*US%L_to_Z**2)
+        call hchksum(MEKE%mom_src_bh, 'MEKE mom_src_bh', G%HI, unscale=US%RZ3_T3_to_W_m2*US%L_to_Z**2)
       if (allocated(MEKE%GME_snk)) &
         call hchksum(MEKE%GME_snk, 'MEKE GME_snk', G%HI, unscale=US%RZ3_T3_to_W_m2*US%L_to_Z**2)
       if (allocated(MEKE%GM_src)) &