"Full cost" method cannot use shadow prices as input

[ahawkes]

Describe the bug

When commodities are "full cost" priced, at present we calculate prices as described here https://energysystemsmodellinglab.github.io/MUSE2/api/muse2/simulation/prices/fn.calculate_full_cost_prices.html

The calculation relies on the input commodities' shadow price. This can result in a "full cost" commodity down stream of another "full cost" commodity having a lower price. This is because shadow prices can be zero or other odd values according to capacities etc (which I'll make another issue out of).

To Reproduce

See attached model. Look at price results for commodity GASPRD and GASNAT. Even in the base year 2020 for the peak.peak time slice, GASNAT is cheaper than GASPRD. This should be impossible because GASPRD is input to a process that produces GASNAT with efficiency <1 and some operating cost and emissions.

Expected behavior

I think we need to calculate the price of each commodity sequentially from top to bottom of the model. In the simple model this would start with GASPRD, then GASNAT etc, (opposite order to the investment appraisal order?) and we would not use shadow prices at all in the "full cost" calculation. I imagine this method will encounter problems when there is a circularity. I don't have a solution to hand for that - thoughts appreciated.

simple.zip

[ahawkes]

Perhaps also related to this, for the GASPRD and GASNAT commodities in the simple model - when using full cost - the price varies between time slices. I think this should not be possible. Variable costs are the same regardless of time slice for these processes, and CAPEX and FIXOM should be distributed according to activity (so per unit annual fixed costs should be the same in each time slice and therefore price should be the same in each time slice)?