[MITgcm-support] Changes in calculation of Visbeck K

[Jean-Michel Campin]

Hello,

Few changes in the way Visbeck eddy diffusivity [Visbeck etal, JPO, 1997]
(VisbeckK, in package gmredi, enable with #define GM_VISBECK_VARIABLE_K
in GMREDI_OPTIONS.h) is computed:
- the averaging over "GM_Visbeck_depth" was slightly wrong and
  has been fixed.
- it no longer depends on the tapering scheme (is also independent from
  "GM_Small_Number"), and just uses the slope (limited by "GM_maxSlope").

The old version has not been removed (need to #define OLD_VISBECK_CALC
at the top of pkg/gmredi/gmredi_calc_tensor.F to recover it) for now,
but will not be maintained.

Cheers,
Jean-Michel