Add "primary output" field to processes

[alexdewar]

This is required for agent investment in the new model.

While we have done away with PACs, for the agent investment we do need to nominate a primary output for every process that produces anything. In the case of processes with one output flow, this will be the primary output, but for processes with multiple outputs the user will need to specify which the primary is. (For processes that only consume, this is obviously not required.)

We should add a primary_output column to processes.csv for this. For processes without outputs, it could be left blank. We could also allow users to leave it blank in the (common) case that there is only one output. In terms of the implementation, it might make sense to store this as a boolean attribute in ProcessFlow for convenience (as we had with PACs before).

[alexdewar]

On second thoughts, I think it might be worth having a column for primary_output in process_flows.csv instead (boolean/empty), because process flows can vary by region and year. So the resulting code is going to look similar to what we had for PACs before.

We could adopt a similar approach to what we had before and sort process flows so that the primary output is at the start of the collection, which would make it easier to look up.

[alexdewar]

We probably could do this now, but I'm going to put it on hold as it isn't obvious where exactly in the code the feature will be used (searching the document for "primary output" yielded nothing) and that may affect the implementation.

[alexdewar]

I'm going to have a go at this now. See Adam's comment here for context: #589 (comment)

It's needed for investment appraisal.