Unstable cases treatment in zdfiwm (formerly zdftmx)

[gsamson]

Context

By default, the unstable cases (bn2 < 0) leads to an estimation of zav_wave of 100 cm2/s.
This treatment is useless in the case of using the TKE standard turbulent scheme as evd will specify a value of 10 or 100 m2/s.
However, this specific treatment can induce some troubles in the case of using a convective model such as the EDMF schemes recently introduced. Indeed, this scheme works in case of instabilities, so they should not not be removed by mixing.

Proposal

The suggestion is to set zav_wave to the background value in unstables cases and let evd, gls or a convective model to deal with.