Started on: making Loki plan trafo (pipeline) aware

[MichaelSt98]

Corresponding CLOUDSC branch: nams-cloudsc-loki-plan (necessary since CLOUDSC wasn't using loki_transform_target yet).

• more or less just copied/utilised the infrastructure and mechanism for (dispatching) transformations and duplicated for planning
• the only trafo that uses this for now is the FileWriteTransformation
  • consequently write_cmake_plan is not renaming the modules/routines anymore

Next step:

• create new item (and/or delete existing item)
  • I've added cloudsc_init.F90 in the above CLOUDSC branch as well as a DuplicateKernel and RemoveKernel trafo to test that (obviously not working yet)

[github-actions]

Documentation for this branch can be viewed at https://sites.ecmwf.int/docs/loki/444/index.html

[reuterbal]

Many thanks for this @MichaelSt98 !

I looked through this yesterday and noticed the need for significant code duplication if we make the planning a separate traversal mode in the Scheduler. So, I've played around with an approach, where planning becomes an option in the Scheduler's process and Transformation's apply entry points, with the latter taking care of dispatching to plan_<...> instead of transform_<...>. For me, that has the advantage that the planning mode really becomes closely linked to the processing of pipelines/transformations.

However, I have not gone as far as testing your DuplicateKernel/RemoveKernel transformations or CLOUDSC in this setup.

You can take a look at the changes here:
main...nabr-pipeline-plan

Note that this sits on top of my changes from #441, because that allowed me to utilize the additional tests added there. I have rebased the branch following the merge, the diff is now clean.

FYI: @mlange05

[MichaelSt98]

Out-dated and superseded due to/by #453.