[MITgcm-support] Changes in diagnostics periodic-averaged binary output file

[Jean-Michel Campin]

Hello MITgcm users,

I've just changed (June 22, shortly after checkpoint62z tag)
the layout of MDS output files (binary output file, not NetCDF)
from pkg/diagnostics when using periodic averaging (averagingFreq,
 averagingPhase and repeatCycle) with multiple fields in the
same file.
The new output file contains: all the fields corresponding to the 1rst 
 time-period, then all the fields from the second, ... etc ...
Whereas the pre-checkpoint62z output file contains
 the first field with all the time-periods, then the
second field with all the time-periods, then the 3rd ... etc ...

To illustrate the case, assuming I want time-average output every 
50.yrs, but keeping the seasonal cycle to produce average over 
each month (12 months) separately (i.e., all January averaged together,
all Feb together ...) of those 3 diagnostic fields:
SI_Fract, SI_Thick and SI_SnowH, the relevant part of data.diagnostics 
will looks like:

> fields(1:3,5) = 'SI_Fract','SI_Thick','SI_SnowH',
>  frequency(5) = 1555200000.,
> averagingFreq(5) = 2592000.,
>   repeatCycle(5) = 12,
>  fileName(5)  = 'iceDiag_12m',

Before checkpoint62z, the MDS output file "iceDiag_12m.[10-digits-iter].*.data"
contains, in this order, 12 months SI_Fract fields + 12 months SI_Thick 
+ 12 months SI_SnowH

Now (after Jun 22), the output file "iceDiag_12m.[10-digits-iter].*.data"
contains 12 sequences (corresponding to the 12 months) 
of the 3 output fields (SI_Fract + SI_Thick + SI_SnowH)

The main motivations for this change are:
a) the time-period record are somehow similar to time dimension
 which is more convenient when pushed in last position, allowing
 to growth the file more naturally by adding other time record.
b) to get the same ordering in NetCDF (MNC output) and binary (MDS output)
 output file (nothing has changed in the NetCDF output which already
 follows the new MDS layout).

Unfortunately, as it is now, there is not much information stored in 
the meta file that would indicate the difference between the pre-checkpoint62z
lay-out and the new one (working to improve this aspect); 
so I would advise to be careful when sorting out time-period and output 
fields ordering in periodic-averaged MDS output file, specially regarding
this pre-checkpoint62z / post-checkpoint62z transition. 

Cheers,
Jean-Michel