Fix submeso scale smoothing

[aidanheerdegen]

Grid scale noise observed in the ACCESS-OM2 models (COSIMA/access-om2#110) could be addressed if the smooth_hblt option was fixed.

!  <DATA NAME="smooth_hblt" TYPE="logical">
!  For smoothing on the submeso bldepth field. This is useful 
!  since the bldepth obtained from KPP or diagnosed mld can 
!  have some grid noise. 
!  Default smooth_hblt=.false. since this agrees with legacy.
!  Note that this scheme fails to reproduce across
!  processor layout, so it remains broken.  
!  </DATA> 

https://github.com/mom-ocean/MOM5/blob/master/src/mom5/ocean_param/lateral/ocean_submesoscale.F90#L1330

[russfiedler]

To expand on this. It appears that the smoothing as it stands allows the the boundary layer depth to be smoothed to below the maximum depth of the water column if the surrounding values of hblt are large enough. It appears that subsequent code assumes that the value of hblt is sensible when calculating kblt etc. There should be a sanity check for the value, something like hblt=min(hblt,zbot) where zbot is the depth of the column.

[StephenGriffies]

I agree @russfiedler ; sanity check is needed.