Support high-order vertical reconstructions for remaping HAMOCC variables to z-levels

[TomasTorsvik]

Physics tracers can now use high-order vertical reconstructions for remapping to z-levels. We should implement this for HAMOCC variables as well.

Ref. #673

[jmaerz]

Hi @TomasTorsvik and @JorgSchwinger , seeing this issue, I just would like to make two comments:

• if it is easy to perform the remapping for iHAMOCC and not too many changes are needed, I would say, we can go ahead
• if the remapping requires significant code change, I would highly suggest to consider rewriting the whole output infrastructure in iHAMOCC. I am not seeing it to be sustainable on the long term the way it is currently written

The latter is about:

• potentially increasing parallelization of output writing for future better performance
• making it easier to write new output variables (ideally having ONE place to define a new variable - including masking, etc.)

It is certainly a bit of a larger task, but potentially worth it... (while ideally, some technically more skilled person should take that on). I am aware that it is likely not fitting the time line for the Fast Track initiative, but just my thoughts on the subject.

[matsbn]

I can just add some experience from the BLOM implementation in #673. For BLOM, the output routines are likely even more messy than iHAMOCC's. They should ideally be rewritten to approach the "wish list" you mention, @jmaerz, and which I agree to. The output routines were kind of OK and more maintainable in their original implementation, only supporting serial I/O (still multiple actions per variable, but less than now). With PnetCDF came the requirement to complete the definition stage before writing variables (a good strategy anyways), messing up the original workflow. When I rewrote the BLOM restart functionality (mod_restart.F90), I at least made an attempt to combine definition/write actions. Some of these ideas could be taken forward, although the diagnostic output is more complex.

For now, to have this available for fast track, I made a pragmatic approach to patch the existing output infrastructure in BLOM.