potentially test undefined values in stpctl and other places...

[smasson]

Context

At least when nn_hls > 1 and in some configurations (at least SAS), stpctl is stoping the model for non-existing errors... so good... :-(

Analysis

in stpctl mpp_min/max and mpp_min/max_loc are done over the whole domain instead of using only the inner domain. In consequence, they sometimes pick-up values (randomly) from non-initialized parts of the halos.

Fix

Make sure that we use only the inner domain when using mpp_min/max and mpp_min/max_loc in stpctl.

Check also the other routines using these functions...

cherry on the cake: modify mpp_min/max_loc so:

• it works even without key_mpp_mpi
• by default, it sents back global domain indices without halos so the information matches input/outputs files and does not depend on nn_hls value.

[smasson]

In 13458:

Error: Failed to load processor CommitTicketReference

No macro or processor named 'CommitTicketReference' found

[smasson]

• This bugfix fixes sette errors with ORCA2_SAS_ICE and ICE_AGRIF when nn_hls = 2.
  
• When nn_hls = 2, we still have sette errors for AGRIF_DEMO, OVERFLOW, LOCK_EXCHANGE...
  
• we close this ticket but further checks are needed for the use (over the whole or the inner domain?) of minval and maxval that are used in 96 routines...