strange behaviour when activating Griffies triads

[cbricaud]

Context

Activating triad operator for active tracers lateral mixing.
Compare with standard iso-neutral operator

Analysis

When comparing 2 ORCA025 simulation ( one with iso-neutral operator , the other with triad opeartor),
In BUG_TRIAD.pdf, some plots show that triad operator slopes are noisy

Fix

In BUG_TRIAD.pdf, some plots show that smoothing triad operator slopes fix the problem

Smoothing with shapiro filter might be added in triad slopes computing as in classical operator

This smoothing will be optional ( add a logical in namelist)

A dedicated subroutine for smoothing will be added and used for classical and triad operator

Will be treated as a development for WP2021; see ticket #2601