Panics if a process end_year is before the end of the simulation

[tsmbland]

Describe the bug

I've just tried adding a milestone year (2040) to the simple example model. This was mostly straightforward and I just had to make sure to add demand data for 2040 (and change the agent objective year to "all" as we were previously specifying it per year).

However, I noticed that if I don't increase the end_year of the processes to 2040 (or beyond) it panics in validate_sed_commodity when trying to get flow data for that year.

To Reproduce

Add 2040 as a milestone year to the simple model

Expected behavior

Ideally we shouldn't panic and the validation should either pass of fail depending on if the model is feasible in that year (it may not be if appropriate processes aren't available for that year), giving a helpful error message if necessary

[tsmbland]

While we're touching this part of the code I think we could consider removing what we've currently done and using the graph library instead.

SVD commodities: need at least one incoming edge (i.e. produced), and no outgoing edges (i.e. not consumed)
SED commodities: any commodity with an outgoing edge (i.e. consumed) must also have an incoming edge (i.e. must also be produced)
Other commodities: incoming or outgoing edges, but not both

We will need to assemble graphs for all milestone years. If we also want to take process availability into account, then we'll need to do it for all timeslices as well

[alexdewar]

@tsmbland Good idea!