standardize FMU contexts

[perolavsvendsen]

fmu.context.stage is (will be) included in the schema definition.

However, this requires some discussion and agreement of the contexts. Current proposal is based on the assumed hierarchy of case/iteration/realization.

• Data produced by an individual realization will have fmu.context.stage: realization. fmu.case, fmu.iteration and fmu.realization is present.
• Data produced on iteration level (e.g. aggregated data across all realizations in an iteration) will have fmu.context.stage: iteration. fmu.case and fmu.iteration will be present.
• Data produced on case level (e.g. observation data produced by pre-processing workflows) will have fmu.context.stage: case. fmu.case will be present.

Tabulated:	fmu.case	fmu.iteration	fmu.realization	fmu.context.stage	Example
✅	❌	❌	"case"	Observation data produced by a pre-process
✅	✅	❌	"iteration"	Aggregated data across an iteration
✅	✅	✅	"realization"	A data object produced within a realization

Note that this has at least two issues:

• It contains the logic twice. E.g. the "iteration" context can be derived either by observing that only fmu.case and fmu.iteration is present, or by reading fmu.context. This further requires synchronization.
• It does not differentiate between "pre-processed" and "post-processed" data on any level. (These terms should also be avoided, as they are relative and can be applied on several levels.)

Specific operation-tags exists, and more could exist. Current example is fmu.aggregation which carries information specifically related to aggregations.

[perolavsvendsen]

Could also imagine the fmu.context.stage to reflect the part of a realization (be more granular), e.g. "structural_model" or "flow_simulation"